Sleep complaints in middle-aged women and men: the contribution of working conditions and work–family conflicts

Authors


Tea Lallukka PhD, Department of Public Health, University of Helsinki, PO Box 41 (Mannerheimintie 172), Helsinki 00014, Finland. Tel: +358 9 191 27566; Fax: +358 9 191 27570; e-mail: tea.lallukka@helsinki.fi

Summary

This study aimed to examine how physical working conditions, psychosocial working conditions and work–family conflicts are associated with sleep complaints, and whether health behaviours explain these associations. We used pooled postal questionnaire surveys collected in 2001–2002 among 40–60-year-old employees of the City of Helsinki (n = 5819, response rate 66%). Participants were classified as having sleep complaints if they reported sleep complaints at least once a week on average (24% of women and 20% of men). Independent variables included environmental work exposures, physical workload, computer work, Karasek’s job strain and work–family conflicts. Age, marital status, occupational class, work arrangements, health behaviours and obesity were adjusted for. Most working conditions were associated strongly with sleep complaints after adjustment for age only. After adjustment for work–family conflicts, the associations somewhat attenuated. Work–family conflicts were also associated strongly with women’s [odds ratio (OR) 5.90; confidence interval (CI) 4.16–8.38] and men’s sleep (OR 2.56; CI 1.34–4.87). The associations remained robust even after controlling for unhealthy behaviours, obesity, health status, depression and medications. Physically strenuous working conditions, psychosocial job strain and work–family conflicts may increase sleep complaints. Efforts to support employees to cope with psychosocial stress and reach a better balance between paid work and family life might reduce sleep complaints. Sleep complaints need to be taken into account in worksite health promotion and occupational health care in order to reduce the burden of poor sleep.

Introduction

Employed people spend about a third of their time at work and an equally long time asleep. Consequently, both these areas are potentially important for health and wellbeing, and are likely to influence each other. Sleep complaints tend to be more prevalent among women, older people, the widowed and divorced, and among those in lower socioeconomic positions (Arber et al., 2009; Ribet and Derriennic, 1999; Roth and Roehrs, 2003; Sekine et al., 2006). Sleep complaints are also connected to various health-related outcomes such as cardiovascular disease (Schwartz et al., 1999; Wolk et al., 2005), sickness absence (Sivertsen et al., 2009) and disability pensions (Sivertsen et al., 2009).

Furthermore, sleep complaints are an increasingly important problem both in the general and the working populations (Metlaine et al., 2005). A previous study suggests that sleep complaints are more prevalent in Finland than in some other European countries (Ohayon and Partinen, 2002). Trends in Finland over the last three decades follow a minor decline in sleep duration, but an increase in sleep complaints (Kronholm et al., 2008). The increase has been greatest among employed middle-aged people, suggesting a need for research on the determinants of poor sleep in this population. The economic burden of sleep complaints among employees is also considerable (Daley et al., 2009; Metlaine et al., 2005). To help employees maintain their ability to work, it is vital to increase our understanding about the work-related and other determinants of sleep complaints.

The conceptual framework of this study is based on sleep complaints and their associations with three key domains of potential determinants among employed people, i.e. a variety of working conditions, the balance between paid work and family life and health behaviours. Health behaviours may be related to both of the other two domains, or act as potential confounders. All these domains of factors are assumed to be linked with sleep complaints, but their mutual effects and associations have not yet been studied. The rationale to focus on these domains is based on their assumed causal effects on sleep. This assumption is in line with a recent four-wave follow-up study that showed causal associations between job demands and job control and sleep complaints (de Lange et al., 2009). However, physical working conditions and work–family conflicts were not addressed and the analysis did not consider the potential confounding role of health behaviours or obesity.

Earlier studies examining work-related factors and sleep complaints or sleep duration have focused typically on shift work and long working hours, both of which have been found to be related to poorer or shorter sleep (Åkerstedt et al., 2007; Dahlgren et al., 2006; Kageyama et al., 2001; Ribet and Derriennic, 1999; Van Dongen, 2006). Sleep complaints are also common among daytime employees (Metlaine et al., 2005). Overall, the prevalence of sleep complaints in the working population has varied around 16–30% (Metlaine et al., 2005; Nakata et al., 2004; Ribet and Derriennic, 1999), whereas the prevalence of more severe chronic insomnia is about 10% among adults (Roth and Roehrs, 2003).

The time exposed to physical working conditions such as uncomfortable working postures, heavy lifting, noise, cold and microbial exposures also affect sleep (Metlaine et al., 2005). A French study reported associations between characteristics of the physical work environment and the incidence of sleep complaints (Ribet and Derriennic, 1999). In particular, exposure to vibration exacerbated sleep complaints.

Several work-related psychosocial factors have been found to be associated with sleep complaints (Aasa et al., 2005; Eriksen et al., 2008; Jansson-Fröjmark et al., 2007; Kivistöet al., 2008; Knudsen et al., 2007; Linton, 2004; Ota et al., 2005; Ribet and Derriennic, 1999). High job demands and low job control have been associated with poorer sleep due to increased arousal during both daytime and night-time (Eriksen et al., 2008; Roth and Roehrs, 2003).

Recently, attention has been paid to conflicts between work and family and their interference with sleep. Both work-to-family and family-to-work conflicts have had adverse effects on sleep quality in Japanese and Swedish populations (Nylén et al., 2007; Sekine et al., 2006). However, there have been no studies of the independent effects of work–family conflicts on both women’s and men’s sleep complaints, taking into account various working conditions and potential confounding factors such as health behaviours. A review focused on psychosocial stress and impaired sleep also concluded that further studies need to examine the combined effects of work- and family-related stress, as few studies have addressed the relative importance of these two domains (Åkerstedt, 2006). Examining work–family conflicts alongside job strain and other working conditions may thus help to shed light on this area.

This study aimed to examine whether sleep complaints are associated with physical working conditions, psychosocial working conditions and work–family conflicts, while also adjusting for health behaviours and obesity.

Methods

Data

The data were derived from the Helsinki Health Study baseline questionnaire surveys conducted in 2001–2002 among employees of the City of Helsinki, aged 40, 45, 50, 55 and 60 years (Lahelma et al., 2005). The response rate was 66% (69% for women, 60% for men). Non-response analyses suggest that the data are broadly representative of the target population (Laaksonen et al., 2008). These data sets comprised 5819 respondents, with data available concerning the variables of interest for 4120 women and 1029 men. The low proportion of men in these data corresponds to the proportion of men who are public sector employees within the City of Helsinki, as well as other municipal workplaces in Finland (Väänänen et al., 2009). There were no differences in the prevalence of sleep complaints between those included and excluded from the analyses because of missing responses on the independent variables. Age, physical working conditions and psychosocial job strain tended to be slightly higher for those excluded from the analyses, while somewhat more manual workers and single/divorced people were among the excluded participants.

Sleep complaints

The four-item Jenkins Sleep Questionnaire (Jenkins et al., 1988) was used to examine self-reported sleep complaints during the previous 4 weeks. The items were ‘having trouble falling asleep’, ‘waking up several times per night’, ‘having trouble staying asleep’ and ‘waking up after the usual amount of sleep feeling tired and worn out’. Six response alternatives ranged from ‘never’ to ‘every night’. Responses to the items were summed, and the summary score was divided by the number of responses. If the score was three or more, indicating sleep complaints at least once a week during the previous month, the respondent was classified as having frequent sleep complaints. Others served as a reference group. The score was calculated for those with responses on at least two of the four items. Data on all four items were completed by 93% of respondents, while 5% had responses on two or three of the four items. Fewer than 2% had only one response, and fewer than 1% of respondents left the whole inventory unanswered. These latter respondents (n = 139, 2.4%) were excluded from the analyses. Similar procedures for calculating sleep complaints have also been followed in other studies (Jerlock et al., 2006; Vahtera et al., 2006).

Work arrangements

Work arrangements included number of working hours and shift work. Working hours were measured on a five-point scale ranging from 1 h to more than 50 h a week. The cutoff point for long working hours was more than 40 h a week. Those working 40 h or less formed the reference category. Current working schedule was categorized as regular daytime work, shiftwork with no night shifts, shiftwork with night shifts including regular nightwork and other working time arrangements.

Working conditions

Physical working conditions were based on factor analysis of an 18-item inventory of environmental and physical exposures at work (Piirainen et al., 2003). A three-factor solution was reached. The first factor comprised ‘work environmental exposures’, such as to hazardous chemicals, climate and noise. The second factor comprised ‘physical workload’, such as uncomfortable postures, repetitive trunk rotation, repetitive movements, standing, lifting and carrying. The third factor comprised ‘working with computer and mouse’ and sitting. Loadings for all three factors were divided into quartiles and included as class variables in all analyses. Other details of the items and the method are reported elsewhere (Laaksonen et al., 2006). In addition, to confirm the results from the factor analysis, we constructed three summed scores of physical working conditions based on the 18 items enquiring about physical working conditions, environmental exposures and work with a computer. The results were similar for both methods, the summed scores and factor analysis, suggesting that these results are not biased or misleading.

Karasek’s job strain model was included as an indicator of psychosocially strenuous work environment (Karasek, 1979; Karasek et al., 1981). According to this model, ‘low job strain’ is the combination of low job demands and high job control, whereas ‘high job strain’ is the combination of high job demands and low job control to surmount the perceived demands. ‘Active work’ is the combination of high job demands and high job control, whereas ‘passive work’ is the combination of low job demands and low control. ‘High job strain’ is expected to have the most detrimental health effects, including poorer sleep (Aasa et al., 2005; Belkic et al., 2004; Chandola et al., 2008; Jansson-Fröjmark et al., 2007; Kivimäki et al., 2006). Nine items on job demands and nine items on job control were included (Karasek, 1985). The responses to all items were summed and the median in the distribution of the summed score was used as a cutoff point for high job demands and high job control to produce the four categories of the job strain model. A full list of the items and further details are reported elsewhere (Lallukka et al., 2006).

Work–family conflicts

Work–family conflicts were examined in order to take into account potential sources of psychosocial influences outside the workplace. An eight-item measure of conflicts between work and family, i.e. whether job responsibilities interfere with family life (four items) and whether family life interferes with job responsibilities (four items), was adopted from the US National Study of Midlife Development (Grzywacz and Marks, 2000). The individual items are shown in Appendix 1 and also reported elsewhere (Roos et al., 2007; Winter et al., 2006). One of the items enquired about family activities and sleep complaints in terms of job performance: ‘Family activities stop you getting the amount of sleep you need to do your job well’. It was omitted initially from the work–family conflict score in this study, but this exclusion did not affect any of the examined associations between working conditions, work–family conflicts and sleep complaints, except that the magnitude of estimates for work–family conflicts was slightly stronger (data not shown). As the differences were practically negligible we retained the item, i.e. used the full original eight-item inventory. Work-to-family and family-to-work conflicts were combined in our study following previous procedures and suggestions (Grzywacz and Marks, 2000; van Hooff et al., 2005). Additional control analyses using the two dimensions of conflicts separately produced similar results (data not shown). Thus both dimensions of work–family conflicts were associated with sleep complaints in a similar way to the combined work–family conflicts measure. The two measures correlated highly with each other (r = 0.63).

Four response alternatives for the eight items were ‘not at all’, ‘to some extent’, ‘a great deal’ and ‘I don’t have a family’. Those who reported that they did not have a family formed a separate category in all the analyses. For all others, responses to the items were summed with scores ranging from 8 to 24. Three groups were formed to compare those with strong (13–24) and those with weak work–family conflicts (9–12) to those with no conflicts (score 8). Control analyses confirmed that the found associations were not sensitive to the chosen cutoff points.

Health behaviours and obesity

Earlier published reports have shown that health behaviours and obesity are associated with sleep complaints (Fogelholm et al., 2007; Nasermoaddeli et al., 2005; Strine and Chapman, 2005). As working conditions also show some associations with health behaviours and obesity (Kouvonen et al., 2005, 2007; Lallukka et al., 2004, 2005, 2008; Siegrist and Rödel, 2006), these were treated as potential confounders of the examined associations. In order to test the robustness of associations between working conditions and sleep, we thus controlled for these factors.

Detailed descriptions of our health behaviour variables have been reported earlier (Lallukka et al., 2008). Smoking was dichotomized to current smokers and non-smokers. Heavy drinking was measured by consuming more than 280 g of pure alcohol per week for men and 140 g per week for women (Working Group appointed by the Finnish Society of Addiction Medicine, 2005). The alcoholic content of beer, wine and spirits was multiplied by the number of units consumed during an average week. Leisure-time physical activity was measured by reported amount and intensity of activities. Metabolic equivalent task (MET) values (Kujala et al., 1998) were calculated, and those in the lowest quintile of the distribution of the MET scores were classified as physically inactive. Healthy food habits were measured based on Finnish national recommendations (National Nutrition Council, 1998) of daily consumption of fruit or berries, fresh vegetables and dark bread, fish at least twice a week, soft vegetable margarine on bread and vegetable oil in cooking. Those reporting two or fewer such healthy food habits were classified as having ‘unhealthy food habits’. Body mass index (BMI = weight/height × height, kg/m2) was based on self-reported height and weight, with obesity measured by a BMI of at least 30.

Covariates

Age was a covariate in all analyses. Marital status was classified into married or cohabiting, single and divorced or widowed. Socioeconomic position was measured by occupational class, categorized into professionals and managers, semi-professionals, routine non-manual employees and manual workers. These covariates were included because of their associations with sleep, working conditions and health behaviours (Arber et al., 2009; Metlaine et al., 2005; Moore et al., 2002; Ribet and Derriennic, 1999; Sekine et al., 2006).

Statistical methods

Distributions of study variables and the prevalence (%) of sleep complaints by these variables are shown in Table 1. Logistic regression analysis was used to examine the associations between sleep complaints and working conditions and work–family conflicts (Tables 2–4). Bivariate models adjusted for age were fitted first (Table 2) to show the individual effects of each exposure variable before further adjustments.

Table 1.   Distributions of study variables and prevalence (%) of sleep complaints by study variables
 Women (%)Men (%)Sleep complaints at least once a week
Women (%)P-valueMen (%)P-value
  1. BMI, body mass index.

Sociodemographic and socioeconomic variables
 40 years201817 23 
 45 years232020 21 
 50 years212127 20 
 55 years252629 18 
 60 years111531<0.0001170.66
 Married and cohabiting677824 18 
 Single131121 24 
 Divorced/widowed1911260.14270.05
 Managers, professionals274526 18 
 Semi-professionals202025 19 
 Routine non-manual421023 22 
 Manual workers1125250.24230.39
Work arrangements
 Regular daytime work797224 19 
 Shiftwork (no nightwork)12928 24 
 Shiftwork (with nightwork)61520 24 
 Other34250.07190.31
 Working 1–40 h a week867624 19 
 Working more than 40 h a week (overtime)1424280.03220.27
Physical working conditions
Environmental exposures – lowest quartile252519 12 
 25 < 50% environmental exposures252522 19 
 50 < 75% environmental exposures252627 20 
 Highest quartile of environmental exposures252429<0.000129<0.0001
Physical workload – lowest quartile252419 17 
 25 < 50% level of physical workload252622 17 
 50 < 75% level of physical workload252523 21 
 Highest quartile of physical workload242433<0.0001250.08
Computer work – lowest quartile242421 25 
 25 < 50% level of computer work252620 14 
 50 < 75% level of computer work262524 15 
 Highest quartile of computer work252533<0.0001260.0006
Psychosocial job strain
 Low job strain252715 12 
 Passive work262922 19 
 Active work262627 21 
 High job strain221834<0.000131<0.0001
Work–family conflicts
 No work–family conflicts13169 12 
 Weak work–family conflicts545421 17 
 Strong work–family conflicts262439 30 
 No family7627<0.000125<0.0001
Health behaviours and obesity
 Non-smoker777324 19 
 Current smoker2327250.59240.06
 Moderate drinker939324 19 
 Heavy drinker77340.0001320.01
 Physically active757323 18 
 Physically inactive2527280.002250.02
 Healthy food habits786224 20 
 Unhealthy food habits2238270.05200.83
 Normal weight868623 19 
 Obese (BMI 30 + )1414290.002250.10
 n (total); % sleep complaints4120102924 20 
Table 2.   Sleep complaints and working conditions among women and men. Logistic regression analysis (age-adjusted bivariate models*)
 Women (= 4120)Men (= 1029)
ORCI 95%ORCI 95%
  1. *The association of age with sleep is shown first. The odds for all the other variables are then adjusted for age

  2. CI, confidence interval; OR, odds ratio.

Sociodemographic and socioeconomic variables
 40 years1.00 1.00 
 45 years1.23(0.96–1.56)0.89(0.55–1.44)
 50 years1.78(1.41–2.26)0.81(0.50–1.31)
 55 years2.05(1.64–2.57)0.75(0.48–1.19)
 60 years2.25(1.72–2.94)0.70(0.41–1.20)
 Married and cohabiting1.00 1.00 
 Single0.87(0.70–1.09)1.35(0.84–2.16)
 Divorced/widowed1.02(0.85–1.22)1.63(1.03–2.57)
 Managers, professionals1.00 1.00 
 Semi-professionals1.03(0.83–1.27)1.06(0.70–1.61)
 Routine non-manual0.86(0.72–1.02)1.21(0.72–2.05)
 Manual workers0.97(0.75–1.24)1.33(0.91–1.95)
Work arrangements
 Regular daytime work1.00 1.00 
 Shiftwork (no nightwork)1.29(1.04–1.61)1.33(0.80–2.22)
 Shiftwork (with nightwork)0.84(0.60–1.16)1.34(0.88–2.04)
 Other1.02(0.67–1.58)0.95(0.43–2.10)
 Working 1–40 h a week1.00 1.00 
 Working more than 40 h a week (overtime)1.21(0.99–1.47)1.21(0.86–1.72)
Physical working conditions
Environmental exposures – lowest quartile1.00 1.00 
 25 < 50% environmental exposures1.16(0.94–1.44)1.68(1.03–2.73)
 50 < 75% environmental exposures1.55(1.25–1.90)1.83(1.13–2.96)
 Highest quartile of environmental exposures1.76(1.43–2.16)2.98(1.87–4.77)
Physical workload – lowest quartile1.00 1.00 
 25 < 50% level of physical workload1.22(0.98–1.51)0.99(0.63–1.56)
 50 < 75% level of physical workload1.32(1.07–1.64)1.27(0.81–1.98)
 Highest quartile of physical workload2.09(1.70–2.56)1.59(1.02–2.46)
Computer work – lowest quartile1.00 1.00 
 25 < 50% level of computer work0.97(0.78–1.21)0.51(0.32–0.80)
 50 < 75% level of computer work1.33(1.08–1.65)0.55(0.35–0.86)
 Highest quartile of computer work1.96(1.60–2.41)1.05(0.70–1.58)
Psychosocial job strain
 Low job strain1.00 1.00 
 Passive work1.58(1.26–1.98)1.61(1.01–2.56)
 Active work2.20(1.77–2.74)1.92(1.20–3.07)
 High job strain3.07(2.46–3.83)3.21(1.99–5.18)
Work–family conflicts    
 No work–family conflicts1.00 1.00 
 Weak3.08(2.24–4.24)1.55(0.92–2.60)
 Strong8.51(6.12–11.82)3.04(1.76–5.23)
 No family3.88(2.59–5.82)2.44(1.16–5.12)
Table 3.   Sleep complaints, working conditions, and work–family conflicts among women (n = 4120)
 Model 1Model 2Model 3Model 4Model 5
ORCI 95%ORCI 95%ORCI 95%ORCI 95%ORCI 95%
  1. Model 1: age, marital status, occupational class, work arrangements, and physical working conditions adjusted for. Model 2: age, marital status, occupational class, work arrangements, and psychosocial working conditions adjusted for. Model 3: age, marital status, occupational class, work arrangements, and work–family conflicts adjusted for. Model 4: age, marital status, occupational class, work arrangements, physical working conditions, psychosocial job strain and work–family conflicts adjusted for. Model 5: age, marital status, occupational class, work arrangements, physical working conditions, psychosocial job strain, work–family conflicts, health behaviours (smoking, alcohol consumption, physical inactivity, unhealthy food habits) and obesity adjusted for.

  2. CI, confidence interval; OR, odds ratio.

Physical working conditions
Environmental exposures – lowest quartile1.00     1.00 1.00 
 25 < 50% environmental exposures1.12(0.90–1.39)    1.02(0.81–1.27)1.02(0.82–1.28)
 50 < 75% environmental exposures1.39(1.12–1.72)    1.19(0.95–1.48)1.19(0.95–1.48)
 Highest quartile of environmental exposures1.57(1.27–1.95)    1.22(0.98–1.53)1.24(0.99–1.55)
Physical workload – lowest quartile1.00     1.00 1.00 
 25 < 50% level of physical workload1.22(0.98–1.53)    1.20(0.95–1.50)1.21(0.96–1.52)
 50 < 75% level of physical workload1.31(1.04–1.64)    1.18(0.93–1.49)1.21(0.95–1.53)
 Highest quartile of physical workload2.31(1.83–2.91)    1.84(1.45–2.34)1.89(1.48–2.40)
Computer work – lowest quartile1.00     1.00 1.00 
 25 < 50% level of computer work1.11(0.87–1.40)    1.03(0.81–1.32)1.03(0.81–1.31)
 50 < 75% level of computer work1.33(1.05–1.68)    1.21(0.95–1.54)1.20(0.94–1.52)
 Highest quartile of computer work2.09(1.66–2.63)    1.56(1.23–1.98)1.55(1.22–1.97)
Psychosocial job strain
 Low job strain  1.00   1.00 1.00 
 Passive work  1.70(1.35–2.15)  1.50(1.18–1.91)1.50(1.17–1.91)
 Active work  2.09(1.67–2.61)  1.40(1.11–1.77)1.41(1.12–1.79)
 High job strain  3.20(2.55–4.01)  1.95(1.53–2.49)1.93(1.51–2.47)
Work–family conflicts
 No work–family conflicts    1.00 1.00 1.00 
 Weak    3.13(2.27–4.32)2.48(1.79–3.45)2.48(1.79–3.44)
 Strong    8.86(6.33–12.40)6.00(4.23–8.50)5.90(4.16–8.38)
 No family    3.77(2.46–5.77)2.90(1.88–4.48)2.89(1.87–4.46)
Model fit statistics
 Max-rescaled R-square0.0824 0.0694 0.1187 0.1537 0.1624 
Table 4.   Sleep complaints, working conditions and work–family conflicts among men (n = 1029)
 Model 1Model 2Model 3Model 4Model 5
ORCI 95%ORCI 95%ORCI 95%ORCI 95%ORCI 95%
  1. Model 1: age, marital status, occupational class, work arrangements and physical working conditions adjusted for. Model 2: age, marital status, occupational class, work arrangements and psychosocial working conditions adjusted for. Model 3: age, marital status, occupational class, work arrangements and work–family conflicts adjusted for. Model 4: age, marital status, occupational class, work arrangements, physical working conditions, psychosocial job strain and work–family conflicts adjusted for. Model 5: age, marital status, occupational class, work arrangements, physical working conditions, psychosocial job strain, work–family conflicts, health behaviours (smoking, alcohol consumption, physical inactivity, unhealthy food habits) and obesity adjusted for. CI: confidence interval; OR: odds ratio.

Physical working conditions
Environmental exposures – lowest quartile1.00     1.00 1.00 
 25 < 50% environmental exposures1.63(0.98–2.69)    1.63(0.98–2.73)1.61(0.96–2.70)
 50 < 75% environmental exposures1.81(1.09–3.01)    1.69(1.00–2.83)1.64(0.97–2.77)
 Highest quartile of environmental exposures3.06(1.76–5.30)    2.47(1.40–4.35)2.42(1.36–4.29)
Physical workload – lowest quartile1.00     1.00 1.00 
 25 < 50% level of physical workload1.04(0.65–1.67)    0.94(0.58–1.52)0.93(0.57–1.52)
 50 < 75% level of physical workload1.12(0.70–1.81)    1.01(0.62–1.64)1.02(0.62–1.66)
 Highest quartile of physical workload1.20(0.74–1.96)    1.02(0.62–1.68)1.03(0.62–1.71)
Computer work – lowest quartile1.00     1.00 1.00 
 25 < 50% level of computer work0.67(0.40–1.12)    0.61(0.36–1.03)0.61(0.36–1.04)
 50 < 75% level of computer work0.59(0.35–0.98)    0.54(0.32–0.90)0.54(0.32–0.92)
 Highest quartile of computer work1.16(0.70–1.92)    0.94(0.55–1.58)0.93(0.55–1.58)
Psychosocial job strain
 Low job strain  1.00   1.00 1.00 
 Passive work  1.42(0.86–2.33)  1.18(0.71–1.97)1.18(0.71–1.96)
 Active work  1.92(1.19–3.10)  1.43(0.85–2.38)1.42(0.85–2.38)
 High job strain  2.96(1.80–4.87)  2.01(1.19–3.41)1.97(1.16–3.35)
Work–family conflicts
 No work–family conflicts    1.00 1.00 1.00 
 Weak    1.84(1.07–3.17)1.50(0.86–2.64)1.52(0.86–2.69)
 Strong    3.80(2.10–6.88)2.60(1.38–4.91)2.56(1.34–4.87)
 No family    1.69(0.76–3.78)1.35(0.59–3.07)1.32(0.57–3.04)
Model fit statistics
 Max-rescaled R-square0.0690 0.0506 0.0551 0.1052 0.1197 

A series of nested models are presented for women (Table 3) and for men (Table 4). Physical working conditions, psychosocial working conditions and work–family conflicts were adjusted for in separate blocks for women and men, while also adjusting for age, occupational class, marital status and work arrangements. All these covariates are included in the models because of their role as potential confounders of the examined associations. First, the three physical working conditions (environmental work exposures, physical workload, computer work) are mutually adjusted for (model 1). In model 2, the focus is on psychosocial job strain, whereas in model 3 work–family conflicts are examined. In model 4, all the above working conditions and work–family conflicts are adjusted simultaneously to examine whether the associations in each area remain after mutual adjustment. Finally, in model 5, health behaviours and obesity are adjusted for to examine whether the associations between sleep complaints, working conditions and work–family conflicts are confounded by these behavioural risk factors.

The variance inflation factor values varied between 1.04 and 1.24, suggesting that multicollinearity is not a problem among the studied working conditions (Morrow-Howell, 1994).

In order to assess further the effects of the adjustments, pseudo-R2 values were calculated (Shtatland et al., 2000). The differences in the likelihood-based R2 values between nested models indicate how the fit of the models improves after adjustments. In Tables 3 and 4, ‘Max-rescaled R-squares’ represent corrected scales with a maximum value of 1.00 (100%). The Hosmer–Lemeshow goodness-of-fit test, concerning the null hypothesis that there is no difference between the observed and predicted values of the outcome, was also computed for the models. The null hypothesis was not rejected, i.e. the models fitted the data adequately.

All the analyses were conducted stratified by gender, as women and men largely work in different jobs with differing working conditions. The importance of showing results separately for women and men has been highlighted in a review of occupational health research, as even in the same jobs the exposures and covariates differ between women and men with subsequent effects on outcome (Niedhammer et al., 2000). In addition, interactions between gender and the exposure variables were tested within the pooled data. The analyses were conducted using the SAS statistical program, version 9.1.3 (SAS Institute Inc., Cary, NC, USA).

Ethics

The Helsinki Health Study protocol was approved by the Ethics Committee of the Department of Public Health and by the Ethics Committee of the health authorities of the City of Helsinki.

Results

The prevalence of sleep complaints was 24% among women and 20% among men aged 40–60 years (Table 1). The complaints were more prevalent in older age groups among women, but not among men. Sleep complaints were more prevalent among non-married than married men, but there was no difference in the prevalence of sleep complaints by marital status among women. Sleep complaints did not differ by occupational class. Among women, sleep complaints were more prevalent among those working overtime, with more environmental exposures, more computer work and psychosocial job strain, as well as among those with work–family conflicts, those who were obese and those with unhealthy behaviours except smoking. Among men, the prevalence of sleep complaints by these variables was fairly similar to that for women, but there were no differences in sleep complaints concerning working overtime, unhealthy food habits or obesity.

All the physical working conditions, psychosocial job strain and work–family conflicts were associated with sleep complaints after adjusting for age only in the logistic regression analyses (Table 2). However, the associations varied by gender. The associations for work environmental exposures were strong among men [odds ratio (OR) 2.98; confidence interval (CI) 1.87–4.77] and women (OR 1.76; CI 1.43–2.16). A high level of physical workload was also associated with sleep complaints both among men (OR 1.59; CI 1.02–2.46) and women (OR 2.09; CI 1.70–2.56), whereas computer work produced differing associations for women and men. While computer work was associated positively with sleep complaints among women (OR 1.96; CI 1.6–2.41), the association among men was curved, with fewer sleep complaints among men in the second (OR 0.51; CI 0.32–0.80) and third (OR 0.55; CI 0.35–0.86) quartiles of extent of computer work. High psychosocial job strain was associated equally with sleep complaints among both men (OR 3.21; CI 1.99–5.18) and women (OR 3.07; CI 2.46–3.83). Strong work–family conflicts were associated with sleep complaints among both men (OR 3.04; CI 1.76–5.23) and women (OR 8.51; CI 6.12–11.82). Even weak work–family conflicts were associated significantly with sleep complaints among women (OR 3.08; CI 2.24–4.24). Both men (OR 2.44; CI 1.16–5.12) and women (OR 3.88; CI 2.59–5.82) with ‘no family’ reported more sleep complaints.

To clarify the inter-relationships between marital status and work–family conflicts we also conducted control analysis by entering only marital status into the model with work–family conflicts (data not shown), but the estimates of the work–family conflicts measure were practically identical to those in Table 2.

In Tables 3 and 4, physical working conditions, psychosocial job strain and work–family conflicts were adjusted for in nested models. Model 1 shows that after mutual adjustments for the three physical working conditions and covariates (age, marital status, occupational class and work arrangements), the associations mostly remained and were even somewhat strengthened. High exposure to ‘work environmental factors’ remained associated strongly with sleep complaints for men (OR 3.06; CI 1.76–5.30), whereas the association was reduced slightly for women (OR 1.57; CI 1.27–1.95). The other physical working conditions, i.e. ‘heavy workload’ and ‘computer and mouse work/sitting’, were associated with sleep complaints among women after adjustments. However, among men, the associations between computer work, physical workload and sleep complaints became non-significant after adjustments (Table 4).

High psychosocial job strain remained associated equally strongly with sleep complaints among both men (OR 2.96; CI 1.80–4.87) and women (OR 3.20; CI 2.55–4.01), after adjustment for age, marital status, work arrangements and occupational class (model 2). The two separate dimensions of the ‘high job strain’ quadrant, i.e. high job demands and low job control, were associated similarly with sleep complaints (data not shown). These separate associations were, however, slightly weaker than those for their combination. Strong work–family conflicts remained associated (model 3) with sleep complaints among men (OR 3.80; CI 2.10–6.88) and women (OR 8.86; CI 6.33–12.40).

When physical working conditions, psychosocial job strain and work–family conflicts (model 4) were adjusted for simultaneously, the strength of the above-reported associations were attenuated somewhat among both genders, but remained statistically significant. Adjusting finally for the four health behaviours and obesity (model 5) had negligible effects on the above results.

The contrasting effects by gender with respect to ‘computer work’ were supported further by control analysis, suggesting an interaction between gender and computer work (P = 0.025) adjusting for age, marital status, occupational class and work arrangements. In relation to the other physical working conditions, psychosocial job strain and work–family conflicts, no statistically significant gender interactions were found (data not shown).

The analyses were repeated using a more severe sleep complaint outcome, i.e. any of the sleep complaints occurring every or almost every night (overall prevalence of 10% in these data, no data shown). The results for physical working conditions, psychosocial job strain and work–family conflicts were practically identical among women (compared with those in Table 3). Among men, most of the associations did not reach statistical significance after adjustments, probably because of the smaller number of men in these surveys.

Discussion

Main findings

This study examined simultaneously the associations of key physical working conditions, psychosocial job strain and work–family conflicts with sleep complaints among middle-aged employees of the City of Helsinki. In addition, the extent to which health behaviours and obesity modified these associations was considered.

The first main finding was that both physical and psychosocial working conditions were related strongly to sleep complaints even after mutual adjustments. In addition, work–family conflicts had independent effects on sleep complaints among both women and men. All these associations were unaffected by occupational class, marital status or work arrangements. However, work–family conflicts partly explained the associations between working conditions and sleep among both women and men. Correspondingly, the effects of work–family conflicts were attenuated when working conditions and psychosocial job strain were adjusted for. The strength and persistence of the associations between working conditions and sleep highlighted that these were independent of health behaviours and obesity. In further analyses (data not shown), we adjusted for self-reported information on doctor-diagnosed pre-existing/lifetime chronic conditions such as cardiovascular diseases, diabetes and depression, current use of medication for hypertension and high cholesterol and menopausal status and hormone therapy among women. These health adjustments had negligible effects on the associations in Tables 3 and 4, confirming the robustness of the associations between working conditions, work–family conflicts and sleep complaints. The associations were also unaffected after adjusting for self-reported sleep duration.

The second main finding was that some differences between women and men were observed concerning the strength of the associations. However, gender interactions were absent except for the effects of computer work on sleep complaints, which was greater for women than men. The effect of psychosocial job strain on sleep complaints was strong among both women and men.

Comparison of current results with previous research

In contrast to previous research about sleep complaints (Arber et al., 2009; Sekine et al., 2006) and sleep duration (Hale, 2005; Krueger and Friedman, 2009), socioeconomic inequalities in sleep complaints were not observed in this cohort. This may be because of a relatively homogeneous study population in terms of age groups (40–60) and employment within a single public sector workplace in the capital area of Finland. In addition, there might be country and culture variations with respect to both socioeconomic differences, working conditions and sleep complaints.

The finding of a lack of consistent association between shift work and sleep complaints also differs from earlier studies (Metlaine et al., 2005). However, our study was based on employees among whom under a quarter undertook shiftwork, with only 6% of women and 15% of men undertaking night shifts. The cross-sectional design and healthy worker effect may partly explain our lack of association (Wilcosky and Wing, 1987). The lack of an association with shiftwork has, nevertheless, also been reported among the general middle-aged Finnish population (Martikainen et al., 2003).

Consistent associations between different physical work environmental exposures and sleep complaints are in line with a previous French study (Ribet and Derriennic, 1999). However, the association between exposure to vibration and incidence of sleep complaints reduced in the French study after adjusting for psychosocial factors, whereas it remained strong in our study after all adjustments. In addition, physical workload was associated fairly strongly with sleep complaints among women. Women who reported the highest level of physical workload were occupied predominantly in healthcare and child daycare. Thus, both jobs and physical exposures related to them are likely to be different between women and men. The sleep complaints of women with a high physical workload may also be related to factors outside work. For example, women who reported the highest level of physical workload and may be too busy to have sufficient sleep time. However, such data were unavailable and thus the reasons underlying current findings need further study.

The contrasting results concerning computer work between women and men may also reflect differences in the nature and repetitiveness of jobs undertaken by women, as well as gender differences in the total time worked with computers. Women, who reported a great deal of computer work, were occupied mainly in health, child and social care and in teaching, whereas men who reported computer work were occupied mainly in teaching and office work.

Although it cannot be excluded in a cross-sectional study that those who have sleep complaints report or perceive their work as more physically strenuous and psychosocially stressful, or that sleep complaints may increase experience of work–family conflicts, reverse causality is unlikely to explain our findings. A previous prospective study examined employees who were free from sleep complaints at baseline, and found an independent association between psychosocial work stress and development of sleep complaints during a 1-year follow-up (Linton, 2004), with the association remaining strong after adjustment for covariates (such as gender, age, health status and working irregular hours). Further confirmation about the causality was found in a recent four-wave follow-up study among Dutch employees, among whom job demands and job control (the dimensions of job strain) and change from a low-strain job to a high-strain job predicted sleep complaints (de Lange et al., 2009). Thus these results, together with ours, suggest that the psychosocial work environment is an important determinant of sleep complaints, and that by lessening the burden of work stress, sleep complaints could potentially be reduced or prevented. In addition, our findings are in line with recent evidence suggesting that those who are frequently bothered or upset at work are more likely to have poor sleep quality, which was not explained by stressful experiences at home (Burgard and Ailshire, 2009).

Strong associations between work–family conflicts and sleep complaints also are in accordance with previous evidence (Jerlock et al., 2006; Nylén et al., 2007; Sekine et al., 2006). Although there were no gender differences in the extent of reported work–family conflicts, one might expect women’s work–family conflicts to be more severe than men’s in terms of their impact on sleep, because of the greater demands of motherhood than fatherhood even in an egalitarian society such as Finland. Women are also particularly likely to worry about their children and other family matters (Arber et al., 2007, 2009), which may partly underlie the effects of work–family conflicts on their sleep.

As health behaviours and obesity did not confound the associations between work–family conflicts and sleep complaints, the mechanisms through which work–family conflicts affect sleep and the work and family-related antecedents of men’s and women’s sleep need further investigation. Reducing stress or helping employees cope with their job demands might reduce sleep complaints that are related to both job strain and work–family conflicts. If strong conflicts between work and family roles are causally and chronically detrimental to sleep, it is vital to also help employees to balance their paid work and family life in order to prevent sleep complaints and subsequent ill-health (Dzaja et al., 2005; Frone et al., 1997; Grant-Vallone and Donaldson, 2001; Hyyppä and Kronholm, 1989; Parish, 2009; Schwartz et al., 1999; Wolk et al., 2005). More research is needed on the direction of the associations between work–family conflicts and sleep, as well as on gender differences with respect to the strength of the associations.

Finally, particular consideration was given to the statistical methods. Recent guidelines advocate adjusted prevalence ratios instead of odds ratios when the outcome is common (Deddens and Petersen, 2008; Tian and Liu, 2006). Accordingly, prevalence ratios were computed using the log-binomial regression models. However, logistic regression models were preferred because they fitted the data better and the odds ratios produced were approximates of prevalence ratios except for work–family conflicts which, nevertheless, remained strong in the results derived using the log-binomial regression as well.

Limitations and strengths

In addition to the above-mentioned limitations related to causal issues and the cross-sectional design, some further limitations are acknowledged. First, health-related selection cannot be excluded (Wilcosky and Wing, 1987). Thus, those with the poorest working conditions and sleep complaints might have exited from the workforce, or might have been less likely to respond to this survey. However, the most robust employees might continue in their strenuous work despite adversities and adverse effects on their sleep. Secondly, this study was limited to public sector employees in Helsinki. The associations may be different in the private sector or in a cohort representative of the entire working population in Finland.

The strengths of this study include a large sample, satisfactory response rate and good quality data. In addition, restricting the study to public sector employees is likely to reduce unmeasured heterogeneity. Moreover, a major advantage of this study was the wide range of variables collected on key working conditions, including work arrangements, physical working conditions and psychosocial job strain, work–family conflicts and covariates. This adds value by enabling different nested models to be examined in order to assess the independence of each set of associations. The opportunity to rule out the effects of health behaviours and obesity as potential confounders is a further strength of this study. In addition, further adjustments for various measures of health status, medication, menopausal status and sleep duration did not alter the reported associations between working conditions, work–family conflicts and sleep complaints.

Conclusions

Working conditions and work–family conflicts showed strong associations with sleep complaints, even after adjusting for a variety of covariates among public sector employees in Finland. Several public health implications should be considered. As sleep complaints are related to ill-health, modification of these examined physical work environment exposures, psychosocial working conditions and work–family conflicts could reduce the risk of chronic diseases attributable to poor sleep. In other words, adverse working conditions may be detrimental to health by contributing to sleep complaints. Furthermore, as sleep complaints are prevalent among middle-aged employees, they need to be taken into account in health promotion programmes at the workplace and in occupational healthcare. Further efforts to help employees’ better cope in psychosocially strenuous work and balance their paid work and family life are also likely to contribute to better sleep. In order to be able to reduce the burden of poor sleep among ageing employees, the causality of the associations between various working conditions and chronic sleep complaints need to be examined further.

Disclosures

None.

Acknowledgements

The Helsinki Health Study is supported by grants from the Academy of Finland (#210435, #205588, #1121748, #130977) and the Finnish Work Environment Fund (#107187, #107281). TL is funded by the Academy of Finland (#133434, #130977), and the Finnish Cultural Foundation, and The Yrjö Jahnsson Foundation. We thank the City of Helsinki, and all members of the Helsinki Health Study group.

Appendix

Appendix 1

Work–family conflicts questionnaire (Grzywacz and Marks, 2000)

First, the work-to-family items asked to what extent job responsibilities interfere with the respondents’ family life.

  • 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.
  • 4 Your job takes so much energy you do not feel up to doing  things that need attention at home.

Correspondingly, the family-to-work items asked the extent to which family life and family responsibilities interfere with the participants’ performance at work.

  • 1 Family matters reduce the time you can devote to your job.
  • 2 Family worries or problems distract you from your work.
  • 3 Family activities stop you getting the amount of sleep you  need to do your job well.
  • 4 Family obligations reduce the time you need to relax or be  yourself.

The response alternatives for all the items were: ‘not at all’, ‘to some extent’, ‘a great deal’ and ‘I don’t have a family’.

Ancillary