International Journal of Behavioral Medicine

, Volume 21, Issue 2, pp 310–318

Conflicts Between Work and Family Life and Subsequent Sleep Problems Among Employees from Finland, Britain, and Japan

Authors

    • Hjelt Institute, Department of Public HealthUniversity of Helsinki
    • Finnish Institute of Occupational Health
  • J. E. Ferrie
    • Department of Epidemiology and Public HealthUniversity College London
    • School of Community and Social MedicineUniversity of Bristol
  • M. Kivimäki
    • Department of Epidemiology and Public HealthUniversity College London
  • M. J. Shipley
    • Department of Epidemiology and Public HealthUniversity College London
  • M. Sekine
    • University of Toyama
  • T. Tatsuse
    • University of Toyama
  • O. Pietiläinen
    • Hjelt Institute, Department of Public HealthUniversity of Helsinki
  • O. Rahkonen
    • Hjelt Institute, Department of Public HealthUniversity of Helsinki
  • M. G. Marmot
    • Department of Epidemiology and Public HealthUniversity College London
  • E. Lahelma
    • Hjelt Institute, Department of Public HealthUniversity of Helsinki
Article

DOI: 10.1007/s12529-013-9301-6

Cite this article as:
Lallukka, T., Ferrie, J.E., Kivimäki, M. et al. Int.J. Behav. Med. (2014) 21: 310. doi:10.1007/s12529-013-9301-6

Abstract

Purpose

Research on the association between family-to-work and work-to-family conflicts and sleep problems is sparse and mostly cross-sectional. We examined these associations prospectively in three occupational cohorts.

Methods

Data were derived from the Finnish Helsinki Health Study (n = 3,881), the British Whitehall II Study (n = 3,998), and the Japanese Civil Servants Study (n = 1,834). Sleep problems were assessed using the Jenkins sleep questionnaire in the Finnish and British cohorts and the Pittsburgh Sleep Quality Index in the Japanese cohort. Family-to-work and work-to-family conflicts measured whether family life interfered with work or vice versa. Age, baseline sleep problems, job strain, and self-rated health were adjusted for in logistic regression analyses.

Results

Adjusted for age and baseline sleep, strong family-to-work conflicts were associated with subsequent sleep problems among Finnish women (OR, 1.33 (95 % CI, 1.02–1.73)) and Japanese employees of both sexes (OR, 7.61 (95 % CI, 1.01–57.2) for women; OR, 1.97 (95 % CI, 1.06–3.66) for men). Strong work-to-family conflicts were associated with subsequent sleep problems in British, Finnish, and Japanese women (OR, 2.36 (95 % CI, 1.42–3.93), 1.62 (95 % CI, 1.20–2.18), and 5.35 (95 % CI, 1.00–28.55), respectively) adjusted for age and baseline sleep problems. In men, this association was seen only in the British cohort (OR, 2.02 (95 % CI, 1.42–2.88)). Adjustments for job strain and self-rated health produced no significant attenuation of these associations.

Conclusion

Family-to-work and work-to-family conflicts predicted subsequent sleep problems among the majority of employees in three occupational cohorts.

Keywords

Sleep QualityWork-family InterfaceProspectiveInternational ComparisonGender

Introduction

Sleep problems are highly prevalent and an increasing number of studies have shown their contribution to various health outcomes including premature mortality [1]. However, less is known about the determinants of sleep problems and prospective studies are rare. There is some evidence that sleep problems have increased over recent decades, particularly among those of working age [2]. Sleep problems are also associated with subsequent work disability among older employees [3, 4]. This highlights the importance of elucidating work-related determinants of sleep problems.

Conflicts between work and family life comprise family-to-work and work-to-family conflicts. As family-to-work and work-to-family conflicts are conceptually two different dimensions [5], they should be examined separately. They may also vary between women and men. Family-to-work conflicts refer to situations where family responsibilities interfere with the abilities to perform at work, while work-to-family conflicts arise if work interferes with family life and the opportunities to fulfill ones demands as a parent or a spouse, for example. Antecedents of family-to-work and work-to-family conflicts have also been shown to vary [6, 7].

In previous studies, conflicts between work and family life have been shown to be associated with mental health problems in particular [79]. In addition, family-to-work and work-to-family conflicts have been found to be associated with sleep problems in cross-sectional studies [1013]. However, measures have varied from single- to multi-item constructs, or the two dimensions have been examined combined. In addition to cross-sectional evidence, family-to-work and work-to-family conflicts have been associated with subsequent use of sleep medication among women [14].

The aim of this study was to examine the associations between family-to-work and work-to-family conflicts at baseline and subsequent sleep problems at follow-up. To increase the generalizability of the results, these associations were examined in three occupational cohorts from Finland, Britain, and Japan composed of women and men in employment at baseline. Additionally, a range of key covariates of the associations were considered.

Methods

Data

We used prospective data from the Finnish Helsinki Health Study [15], the British Whitehall II Study [16], and the Japanese Civil Servants Study [17].

The baseline survey of the Helsinki Health Study was conducted in 2000–2002 among 40- to 60-year-old employees of the City of Helsinki (n = 8,960, response rate 67 %). The follow-up survey was sent in 2007 to all baseline respondents (n = 7,332, response rate 83 %). We included in the analyses all those who had responded at baseline and follow-up, with the exception that blue-collar employees were excluded to increase comparability to the British cohort. After also excluding those without a family, i.e., those to whom the questions on conflicts between work and family life were not applicable (please see a more detailed description of the measures here below), the final study population comprised 3,240 women and 641 men.

The Whitehall II Study is a cohort of white-collar civil servants from 20 London-based civil service departments. At baseline (1985–1988), all participants were aged 35–55 (n = 10,308; response rate, 73 %). The follow-up surveys have been conducted at 2.5-year intervals. To increase comparability to the Finnish and Japanese cohorts, we used phase 5 survey (1997–1999, response rate 76 %) as baseline and phase 7 (2003–2004, response rate 68 %) as follow-up. At phase 5, only the participants who had remained in employment were included (65 %) to reflect the fact that all participants were employed at baseline in the other two cohorts. The analyses included those participants employed at phase 5 who had also returned a questionnaire at phase 7. After excluding those to whom the questions on conflicts between work and family life were not applicable, the final study population comprised 969 women and 3,029 men.

Baseline data for the Japanese Civil Servants Study were collected in 1998–1999. The cohort comprised local government employees aged 20–65 years working in a west coast province of Japan (n = 4,933; response rate, 81 %). The follow-up (2003) included all those working at the time of the survey, irrespective whether they had participated in the baseline survey (n = 4,272; response rate, 79 %). We included in the analyses only white-collar employees, who were participants in both baseline and follow-up surveys and were aged 30 years or over at baseline. Blue-collar workers were excluded as in the Finnish cohort. After excluding those to whom the questions on conflicts between work and family life were not applicable, the final study population comprised 572 women and 1,262 men.

Ethical approvals have been received from the Department of Public Health, University of Helsinki, and the City of Helsinki health authorities (Helsinki Health Study and the University College London) ethics committee (Whitehall II Study), while a committee of civil servants checked the contents and ethical aspects of the Japanese Civil Servant Study. Institutional ethical approval was not required for an observational study in Japan at the time of the survey.

Measures

Conflicts Between Work and Family Life

The measurement of conflicts between work and family life included, firstly, whether family responsibilities interfered with work (family-to-work conflicts, three items), and secondly, whether job responsibilities interfered with family life (work-to-family conflicts, four items). These measures were derived from the US National Study of Midlife Development [5]. Family-to-work conflicts were measured by the following questions: (1) Family matters reduce the time you can devote to your job, (2) family worries or problems distract you from your work, and (3) family obligations reduce the time you need to relax or be by yourself. An additional item inquired whether family activities stopped the respondents from getting the amount of sleep they needed to do their job well. As this item was related to the outcome, it was excluded from the measure.

Work-to-family conflicts were measured by the following questions: (1) your job reduces the amount of time you can spend with the family, (2) problems at work make you irritable at home, (3) your work involves a lot of travel away from home, and (4) your job takes so much energy you do not feel up to doing things that need attention at home. Four response alternatives were included for each question: “not at all”, “to some extent”, “a great deal”, and “I don’t have a family” (Helsinki Health Study)/“not applicable” (Whitehall II Study and Japanese Civil Servants Study). As conflicts between work and family life are only relevant to those who are working and have a family, those who did not have a family or reported that the questions were not applicable to them were excluded from the final study populations analyzed (6 % in the Finnish cohort, 9 % in the British cohort, and 29 % in the Japanese cohort). Other response alternatives were scored from 1 to 3, and the sum scores classified into no, weak, and strong family-to-work and work-to-family conflicts. In each cohort, cut-off points for none, weak, and strong family-to-work conflicts were 3, 4–5, and >5 and for work-to-family conflicts 4, 5–7, and >7, respectively.

Cronbach’s alpha for family-to-work conflicts was around 0.7 in the British and Finnish cohorts and around 0.8 in the Japanese cohorts. The figures for work-to-family conflicts were slightly lower, around 0.6 and 0.7, respectively. Further psychometric properties of the measure have been reported in an earlier comparative paper within the examined cohorts [18]. A relatively small proportion reported both strong family-to-work and work-to-family conflicts in the Finnish and British cohorts while around a third of Japanese participants reported strong conflicts on both dimensions.

Sleep Problems

Sleep problems were measured by the validated Jenkins sleep questionnaire in the Finnish and British cohorts [19] and by the Pittsburgh Sleep Quality Index in the Japanese cohort [20]. The Jenkins questionnaire comprises four items inquiring whether participants have had trouble with initiating sleep, maintaining sleep, waking up too early, and nonrestorative sleep in the previous 4 weeks. As previously [3], those who reported that they had suffered from any of these sleep problems on 15 nights or more in the last 4 weeks were classified as frequent sleep problem cases. Others served as the noncase group.

The Pittsburgh Sleep Quality Index (17 items) was used to produce a global score reflecting the quality of sleep and was based on the sum of seven component scores. Each component score ranges between 0 and 3, and the global score between 0 and 21. Higher scores indicate poorer sleep quality. Following previous procedures [11], a score of more than 5 identified cases of poor sleep quality while lower scores were included in the noncase group. The measure has been shown to be a valid and reliable tool to assess sleep quality [20, 21]. Further details and the Japanese version of the questionnaire can be found elsewhere [11].

The Jenkins questionnaire was available at both baseline and follow-up for the Finnish and British cohorts. The Pittsburgh Sleep Quality Index was only available at follow-up for the Japanese cohort, although a measure of baseline sleep quality was available through a single item question in which the respondents were asked to report some general symptoms in the last 14 days. One of the items asked about “difficulty in sleeping”. The response alternatives were “yes” and “no”. This item was used in the analyses to control for baseline sleep in the Japanese cohort.

Covariates

Covariates included age, sex, job strain, and self-rated health as they are potential confounders or mediators of the association between family-to-work and work-to-family conflicts and sleep problems. Job strain was measured as psychosocial job demands and job control [22]. Those with low job demands (below median score in job demands scale) and high job control (above median score in the job control scale), a combination called “low job strain” were used as the reference and compared with those with active work (high demands and high control), passive work (low demands and low control), or high job strain (high demands and low control). Self-rated health was measured by a single-item asking participants to rate their health from very poor to excellent (5-point scale). Self-rated health is a widely used indicator of health and a powerful predictor of morbidity and mortality [23]. For the analyses, self-rated health was dichotomized following previous procedures to differentiate those in poor health from those with average, good, or excellent health [24].

Statistical Analyses

We first examined age-adjusted prevalence of sleep problems at follow-up according to family-to-work and work-to-family conflicts at baseline, separately in women and men. These associations were then examined in more detail using logistic regression analysis and were initially adjusted for age (model 1) and baseline sleep problems (model 2). Further models additionally adjusted for job strain (model 3) and self-rated health (model 4). We used multiple imputation analyses to control for missing values. The imputation analysis was conducted using the aregImpute function in the Hmisc package [25] for R software (R Foundation for Statistical Computing, Vienna). During the imputation, the procedure created ten imputed datasets, assuming missingness at random. The estimates are obtained by averaging across the results from each of these ten datasets using Rubin’s rules. Sensitivity analyses were conducted using a SAS (version 9.2) program.

Results

Descriptive Results

Work-to-family conflicts were more prevalent than family-to-work conflicts in the Finnish and British cohorts (Table 1) while family-to-work conflicts were more prevalent in the Japanese cohort. Sleep problems were somewhat more prevalent in the British than the Finnish cohort, more prevalent in women than men, and increased over follow-up in both cohorts. In the Japanese cohort, sleep problems between the two time points cannot be directly compared since different measures were used. However, as in the other cohorts, sleep problems were more prevalent among women than men. The age-adjusted prevalence of sleep problems at follow-up by work-to-family conflicts at baseline was similar in all the cohorts with the prevalence of poor sleep generally increasing with the level of conflict (Table 2).
Table 1

Distribution (in percent) of family-to-work and work-to-family conflicts at baseline, and sleep problems at baseline and follow-up among women and men: the Finnish Helsinki Health Study, the British Whitehall II Study, and the Japanese Civil Servants Study

 

Helsinki Health Study

Whitehall II Study

Japanese Civil Servants Study

Women (n = 3,240; %)

Men (n = 641; %)

Women (n = 969; %)

Men (n = 3,029; %)

Women (n = 572; %)

Men (n = 1,262; %)

Family-to-work conflicts at baseline

No

50

54

42

37

5

15

Weak

37

33

38

42

22

27

Strong

13

13

20

21

73

58

Work-to-family conflicts at baseline

No

18

21

19

12

5

7

Weak

65

60

56

54

44

49

Strong

18

19

25

35

51

45

Sleep problems at baselinea

21

15

31

22

11

10

Sleep problems at follow-up

27

19

34

26

32

18

aIn the Japanese Civil Servants Study, a single item sleep measure at baseline and the Pittsburgh Sleep Quality Index at follow-up

Table 2

Prevalence of sleep problems at follow-up by family-to-work and work-to-family conflicts at baseline (age adjusted)

 

Helsinki Health Study

Whitehall II Study

Japanese Civil Servants Study

Women (n = 3,240)

Men (n = 641)

Women (n = 969)

Men (n = 3,029)

Women (n = 572)

Men (n = 1,262)

% (95 % CI)

% (95 % CI)

% (95 % CI)

% (95 % CI)

% (95 % CI)

% (95 % CI)

Family-to-work conflicts

No

24 (22–26)

18 (14–23)

33 (28–38)

22 (19–24)

5 (0–26)

12 (6–19)

Weak

28 (26–31)

23 (17–29)

34 (29–39)

26 (24–29)

32 (22–43)

14 (10–19)

Strong

33 (28–37)

17 (10–27)

40 (33–48)

31 (27–35)

33 (28–39)

22 (18–26)

Work-to-family conflicts

No

20 (17–24)

15 (9–22)

19 (13–26)

17 (13–22)

6 (0–27)

14 (6–26)

Weak

25 (23–27)

20 (16–25)

36 (32–41)

23 (21–25)

23 (16–30)

14 (11–17)

Strong

38 (34–42)

22 (15–30)

43 (36–50)

32 (29–35)

42 (35–49)

24 (20–29)

The Associations Between Family-to-Work and Work-to-Family Conflicts and Subsequent Sleep Problems

In age-adjusted analyses strong family-to-work conflicts were associated with subsequent sleep problems in Finnish women and British men (Table 3). After adjustment for baseline sleep problems (model 2) these associations were considerably attenuated, but strong family-to-work conflicts remained associated with sleep problems among Finnish women (OR, 1.33 (95 % CI, 1.02–1.73)). Adjustment for job strain and health status slightly attenuated these associations. Strong family-to-work conflicts were also associated with sleep problems among Japanese women and men after adjusting for age and baseline sleep problems (OR, 7.61 (95 % CI, 1.01–57.23) and OR 1.97 (95 % CI, 1.06–3.66), respectively). In women these associations were based on very small numbers in the reference category, resulting in very wide confidence intervals. These associations were slightly attenuated after adjustment for job strain or health status (models 3 and 4, women) and self-rated health (model 4, men).
Table 3

Family-to-work conflicts at baseline and subsequent sleep problems at follow-up

 

Model 1a

Model 2b

Model 3c

Model 4d

OR (95 % CI)

OR (95 % CI)

OR (95 % CI)

OR (95 % CI)

Women

 Helsinki Health Study, Finland (n = 3,240)

 No

1.00

1.00

1.00

1.00

 Weak

1.34 (1.12–1.61)

1.17 (0.97–1.41)

1.14 (0.94–1.38)

1.13 (0.94–1.37)

 Strong

1.71 (1.34–2.19)

1.33 (1.02–1.73)

1.27 (0.97–1.66)

1.26 (0.96–1.64)

 Whitehall II Study, Britain (n = 969)

 No

1.00

1.00

1.00

1.00

 Weak

1.03 (0.74–1.43)

0.97 (0.68–1.37)

0.94 (0.66–1.34)

0.96 (0.68–1.36)

 Strong

1.35 (0.91–2.00)

1.13 (0.75–1.72)

1.07 (0.70–1.64)

1.13 (0.75–1.72)

 Japanese Civil Servants Study (n = 572)

 No

1.00

1.00

1.00

1.00

 Weak

7.11 (0.94–53.55)

6.92 (0.89–54.09)

6.50 (0.84–50.14)

7.05 (0.89–55.74)

 Strong

7.79 (1.07–56.48)

7.61 (1.01–57.23)

6.97 (0.93–52.38)

7.43 (0.98–56.45)

Men

 Helsinki Health Study, Finland (n = 641)

 No

1.00

1.00

1.00

1.00

 Weak

1.43 (0.91–2.25)

1.26 (0.77–2.05)

1.23 (0.75–2.01)

1.19 (0.72–1.95)

 Strong

0.98 (0.51–1.88)

0.78 (0.38–1.56)

0.77 (0.38–1.57)

0.67 (0.33–1.37)

 Whitehall II Study, Britain (n = 3,029)

 No

1.00

1.00

1.00

1.00

 Weak

1.31 (1.07–1.61)

1.06 (0.85–1.32)

1.03 (0.82–1.29)

1.05 (0.84–1.31)

 Strong

1.71 (1.34–2.18)

1.26 (0.96–1.64)

1.20 (0.92–1.57)

1.22 (0.94–1.60)

 Japanese Civil Servants Study (n = 1262)

 No

1.00

1.00

1.00

1.00

 Weak

1.20 (0.60–2.39)

1.17 (0.58–2.37)

1.17 (0.57–2.40)

1.13 (0.56–2.29)

 Strong

1.98 (1.08–3.64)

1.97 (1.06–3.66)

1.93 (1.03–3.61)

1.84 (0.99–3.41)

The British Whitehall II Study, the Finnish Helsinki Health Study, and the Japanese Civil Servants Study. Logistic regression analyses, odds ratios (OR), and their 95 % confidence intervals (95 % CI)

aAge adjusted for

bAge + baseline sleep problems adjusted for

cModel 2 + job strain adjusted for

dModel 2 + self-rated health adjusted for

Age-adjusted associations between work-to-family conflicts and sleep problems in women were strong and consistent in all three cohorts (Table 4) and remained after adjustment for baseline sleep problems (OR, 1.62 (95 % CI, 1.20–2.18), 2.36 (95 % CI, 1.42–3.93), and 5.35 (95 % CI, 1.00–28.55) in Finnish, British, and Japanese women, respectively). Further adjustments for job strain and self-rated health (models 3–4) only slightly attenuated these associations. Among men, an association between strong work-to-family conflicts and subsequent sleep problems were seen only in the British cohort after adjustment for age and baseline sleep problems (OR, 2.02 (95 % CI, 1.42–2.88)). This association was little affected by further adjustment for job strain and self-rated health. Finally, we also conducted analyses adjusting for all the covariates simultaneously (data not shown). Our findings remained similar in the British cohort but were attenuated in the Finnish and Japanese cohorts.
Table 4

Work-to-family conflicts at baseline and subsequent sleep problems at follow-up

 

Model 1a

Model 2b

Model 3c

Model 4d

OR

95 % CI

OR

95 % CI

OR

95 % CI

OR

95 % CI

Women

 Helsinki Health Study, Finland (n = 3,240)

 No

1.00

1.00

1.00

1.00

 Weak

1.39 (1.09–1.76)

1.17 (0.92–1.50)

1.09 (0.85–1.41)

1.11 (0.87–1.42)

 Strong

2.48 (1.88–3.29)

1.62 (1.20–2.18)

1.42 (1.04–1.95)

1.44 (1.06–1.95)

 Whitehall II Study, Britain (n = 969)

 No

1.00

1.00

1.00

1.00

 Weak

2.43 (1.56–3.78)

2.12 (1.34–3.34)

2.04 (1.28–3.25)

2.11 (1.34–3.34)

 Strong

3.13 (1.92–5.09)

2.36 (1.42–3.93)

2.23 (1.29–3.85)

2.35 (1.41–3.92)

 Japanese Civil Servants Study (n = 572)

 No

1.00

1.00

1.00

1.00

 Weak

2.88 (0.53–15.71)

2.64 (0.48–14.61)

2.41 (0.43–13.45)

2.64 (0.47–14.87)

 Strong

6.48 (1.24–33.95)

5.35 (1.00–28.55)

4.66 (0.85–25.43)

5.21 (0.96–28.29)

Men

 Helsinki Health Study, Finland (n = 641)

 No

1.00

1.00

1.00

1.00

 Weak

1.48 (0.86–2.58)

1.42 (0.79–2.54)

1.40 (0.77–2.53)

1.36 (0.76–2.46)

 Strong

1.65 (0.85–3.17)

1.34 (0.66–2.71)

1.32 (0.62–2.79)

1.15 (0.56–2.37)

 Whitehall II Study, Britain (n = 3,029)

 No

1.00

1.00

1.00

1.00

 Weak

1.50 (1.09–2.07)

1.45 (1.03–2.05)

1.42 (1.00–2.01)

1.44 (1.02–2.02)

 Strong

2.46 (1.77–3.42)

2.02 (1.42–2.88)

1.87 (1.29–2.71)

1.95 (1.37–2.79)

 Japanese Civil Servants Study (n = 1,262)

 No

1.00

1.00

1.00

1.00

 Weak

1.01 (0.44–2.34)

1.03 (0.44–2.39)

1.07 (0.46–2.50)

1.04 (0.44–2.43)

 Strong

1.98 (0.86–4.57)

1.86 (0.80–4.37)

1.97 (0.83–4.70)

1.79 (0.76–4.25)

The British Whitehall II Study, the Finnish Helsinki Health Study, and the Japanese Civil Servants Study. Logistic regression analyses, odds ratios (OR), and their 95 % confidence intervals (95 % CI)

aAge adjusted for

bAge + baseline sleep problems adjusted for

cModel 2 + job strain adjusted for

dModel 2 + self-rated health adjusted for

Sensitivity Analyses

We have conducted several sensitivity analyses to confirm our results. First, we conducted sensitivity analyses including all four original family-to-work items. The results remained largely similar to those reported in which a sleep-related item was omitted (data not shown). Second, we excluded all those with high job strain at baseline. Although somewhat attenuated, particularly in the Japanese cohort, the associations again remained very similar to those reported. We also stratified our analyses by job strain to examine whether it moderated the associations. As some of the groups (those with high strain) were small, we were unable to draw strong conclusions, however, broadly similar patterns of associations were observed. A third set of sensitivity analyses were conducted in a sample stratified by baseline sleep problems and in samples excluding all those who reported sleep problems at baseline. As the samples reporting baseline sleep problems were smaller, the models were more unstable and confidence intervals wider. However, our results remained similar to those in the full sample, when the analyses were restricted to those not reporting sleep problems at baseline. While all participants were continuously employed in the Japanese cohort, around 30 % of the British and Finnish participants retired or exited the labor market during follow-up. A fourth set of sensitivity analyses, conducted in a sample restricted to those employed at both time points in the British and Finnish cohorts, produced similar results to those reported. Finally, sensitivity analyses examining complete cases produced results similar to those with imputed data (data not shown).

Discussion

Main Findings

This study sought to examine the associations between family-to-work and work-to-family conflicts and subsequent sleep problems in three different occupational cohorts. Our findings provided evidence of age-independent associations between family-to-work conflicts and sleep problems among Finnish women, British men and Japanese employees of both sexes. Work-to-family conflicts were consistently associated with sleep problems in British, Finnish and Japanese women but in men this association was seen only in the British cohort. Sleep problems at baseline made the strongest contribution to these associations, while adjustment for job strain and health status generally had minor effects.

Interpretation

The prospective associations observed between family-to-work and work-to-family conflicts and subsequent sleep problems in the present study confirm earlier cross-sectional evidence [1012] and add to evidence on direction of causality. However, the evidence for an association between family-to-work conflicts and sleep problems was not consistent across the sexes and cohorts and generally did not survive the adjustments. This differs from an earlier study from Finland suggesting that family-to-work conflicts are strongly associated with subsequent sleep medication among women but not among men [14]. As sleep medication was based on prospective register data on prescribed reimbursed medication in that study, it is likely to reflect more clinical and severe sleep problems than those in the present study.

Adjustments for baseline sleep only partially attenuated the associations between family-to-work and work-to-family conflicts and sleep problems, suggesting that, despite the persistence of sleep problems [26, 27], such conflicts bear significance for subsequent sleep problems among middle-aged women and men. That these associations remained after adjustment for job strain and poor health, which have both been associated previously with sleep problems [10, 28, 29], suggests that our findings are not attributable to adverse work characteristics or existing morbidity. These above assumptions are further supported by sensitivity analyses which showed that the associations were found even after excluding those with high job strain and baseline sleep problems.

The analyses in Helsinki men showed weak or no associations between family-to-work and work-to-family conflicts and sleep problems. However, the wide confidence intervals on these effect sizes show that there was little evidence to suggest that these patterns were different to those in the other cohorts. Previously reported cross-sectional associations between a combined measure of family-to-work and work-to-family conflicts and sleep problems were also strong among the Finnish men and remained after all adjustments [10]. It is not known whether a larger sample size would have confirmed the association among Finnish men as well or whether it would have confirmed that men’s sleep problems in the Finnish cohort are unrelated to family-to-work and work-to-family conflicts. Although family and work policy and legislation differ between the countries and the meaning and cultural significance of work varies there is some consistency in our findings, in particular in relation work-to-family conflicts in women. Certainly stronger effects were observed in Japanese women, but work has traditionally been viewed as very important in Japan, and overtime work is common [30]. The greater significance attached to work in Japan could explain the higher prevalence of conflicts, and also the stronger effects on sleep. As attachment to labor market is weaker among Japanese women than men, this might explain the stronger effects of conflicts between work and family perceived among the employed women. Alternatively, differences in effect size could be due to differences in the measurements of sleep between the studies, unmeasured third factors, or small frequencies in some cells.

Methodological Considerations

This study had some limitations. First, the measures of sleep problems differed slightly between cohorts and only a proxy single item measure of sleep quality at baseline was available for the Japanese cohort. Although we cannot rule out that different sleep measures might have affected the findings, strict comparisons between the countries were not made. Thus it is unlikely that this distorted the findings as both measures capture sleep quality and similar insomnia symptoms. Second, study designs and data collection differed slightly between the studies. Baseline and the first follow-up data were used from the Finnish and Japanese cohorts, whereas data collected at phase 5 were used as the baseline in the British cohort and phase 7 as the follow-up time point. However, the data have remained broadly representative of the target population and attrition is unlikely to have strongly distorted the findings [16, 31, 32]. Third, it is possible that other changes have occurred between baseline and follow-up that affected the emergent sleep problems or their fluctuation over time. Fourth, a larger number of participants were excluded from the Japanese cohort due to responding ‘not applicable’ to one or more of the family-to-work and work-to-family conflict items. We lack data to elaborate reasons for this difference between cohorts, but as these participants tended to report more sleep problems, their exclusion is likely to result in conservative estimates of effects. It is also possible that answering “not applicable” is culturally more acceptable than, e.g., nonesponse. Fifth, as the years over which the studies were carried out were not fully comparable between cohorts we cannot rule out the possibility of changes in work and family related policies or culture between the periods. However, as policy and cultural changes are generally incremental and slow to take effect, especially in the Civil Service, we feel that this is an unlikely alternative explanation of our findings. Despite all of these issues, some similarity in the broad patterns of the examined association suggests common effects of conflicts between work and family life to sleep across various contexts and over time.

There are several strengths in this study. The first is a prospective design in a field where the sparse previous studies have been mostly cross-sectional [10, 11]. Second, conflicts between work and family life were measured similarly in each cohort, enabling comparison of their effects on sleep in the three cohorts. Third, similar data on covariates were collected in each cohort and so we were able to rule out the effects of these factors on the associations examined.

Conclusions

Our study provides some evidence that family-to-work conflicts result in subsequent sleep problems and consistent evidence of an association between work-to-family conflicts and sleep problems in British, Finnish, and Japanese women. Conflicts between work and family life have been recognized as an important contributor to mental and physical well-being. Our study suggests that policies that reduce particularly work-to-family conflict may reduce sleep problems among middle-aged workers.

Acknowledgments

The Helsinki Health Study is supported by the Academy of Finland (grants No. 1129225, 1121748, and 1257362) and the Finnish Work Environment Fund (No. 107281). TL and EL have been supported by the Academy of Finland (grants No. 1133434 and 1135630). The City of Helsinki is also acknowledged. MK is supported by the Academy of Finland, the Finnish Work Environment Foundation, the New OSH ERA research program, and a ESRC professorship, UK. The Whitehall II Study has been supported by the British Medical Research Council (MRC); the British Heart Foundation; the British Health and Safety Executive; the British Department of Health; the National Heart, Lung, and Blood Institute (grant No. R01HL036310); the National Institute on Aging (grants No. R01AG013196 and R01AG034454); the Agency for Health Care Policy and Research (grant No. HS06516); and the John D. and Catherine T. MacArthur Foundation Research Networks on Successful Midlife Development and Socioeconomic Status and Health. MK is supported by the Academy of Finland, the Bupa Foundation, and the New OSH ERA (New and Emerging Risks in Occupational Safety and Health, European Research Area) research program and MJS by the British Heart Foundation. The Japanese Civil Servants Study is supported by the Ministry of Health, Labour and Welfare, the Japanese Society for the Promotion of Science, the Occupational Health Promotion Foundation, the Universe Foundation (98.04.017), the Daiwa Anglo-Japanese Foundation (03/2059), and the Great Britain Sasakawa Foundation (2551).

Copyright information

© International Society of Behavioral Medicine 2013