You are here

Article Text

Article menu

Download PDFPDF

Workplace
Original article
Mediating pathways and gender differences between shift work and subjective cognitive function
  1. Imelda S Wong1,
  2. Peter M Smith2,3,
  3. Selahadin Ibrahim2,
  4. Cameron A Mustard2,3,
  5. Monique A M Gignac2,3
  1. 1National Institute for Occupational Safety and Health, Cincinnati, Ohio, USA
  2. 2Institute for Work and Health, Toronto, Ontario, Canada
  3. 3Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada
  1. Correspondence to Dr Imelda S Wong, Division of Applied Research and Technology, National Institute for Occupational Safety and Health, 1150 Tusculum Ave, MS C-24, Cincinnati, OH 452226, USA; IWong{at}cdc.gov

Abstract

Objectives Increased injury risk among shift workers is often attributed to cognitive function deficits that come about as a result of sleep disruptions. However, little is known about the intermediate influences of other factors (eg, work stress, health) which may affect this relationship. In addition, gender differences in these the complex relationships have not been fully explored. The purpose of this study is to (1) identify the extent to which work and non-work factors mediate the relationship between shift work, sleep and subsequent subjective cognitive function; and (2) determine if the mediating pathways differ for men and women.

Methods Data from the 2010 National Population Health Survey was used to create a cross-sectional sample of 4255 employed Canadians. Using path modelling, we examined the direct and indirect relationships between shift work, sleep duration, sleep quality and subjective cognitive function. Multigroup analyses tested for significantly different pathways between men and women. Potential confounding effects of age and self-reported health and potential mediating effects of work stress were simultaneously examined.

Results Work stress and sleep quality significantly mediated the effects of shift work on cognition. Age and health confounded the relationship between sleep quality and subjective cognition. No differences were found between men and women.

Conclusions Occupational health and safety programmes are needed to address stress and health factors, in addition to sleep hygiene, to effectively address cognitive function among shift workers.

  • path analyses
  • subjective cognitive function

https://doi.org/10.1136/oemed-2016-103774

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    What this paper adds

    • Cognitive function deficits among shift workers is often attributed to sleep disruption; however, little is known about the intermediate influences of other factors such as age, health and work stress.

    • Path analysis modelling provided a better understanding of the complex relationship between direct and indirect effects of shift work, sleep and subjective cognition.

    • This study found that greater work stress was a significant mediator in the relationship between shift work and lower sleep quality.

    • To reduce the effect of cognitive function deficits among shift workers, occupational health and safety efforts are needed to address a range of factors such as work stress and health, in addition to sleep disruption.

    Background

    The increased risk of work injury among shift workers has been well established1–3 and is typically attributed to sleep-related fatigue and deficits in cognitive function.4–6 Non-standard schedules, defined as those outside of the hours of 0700–1800 hour,7 can require workers to be awake during conventional sleeping hours, and asleep during regular waking hours. The misalignment of wake/sleep and light/dark cycles can lead to disruption of circadian rhythms and the homoeostatic drive for sleep.8 In addition, cognitive processes follow a distinct diurnal rhythm modulated by the internal circadian clock. When wakefulness occurs at inappropriate biological times, such as with shift work, the misalignment between circadian and wakefulness-sleep physiology can lead to impaired cognitive performance.9 Reduced sleep can alter normal functioning of sustained attention networks in the brain and increases disengagement from external sensory triggers.10 Studies have shown that after only one night shift, workers can exhibit significant sleep impairment and cognitive deficits such as short-term memory.11–13 Over the long term, shift work and chronic sleep deprivation can have lasting effects. Experimental studies have demonstrated that restricting sleep to <6 hours per day for up to 2 weeks in a laboratory setting results in the same cognitive function deficits as would be achieved with 48 hours of total sleep loss.14 Most recently, prospective cohort studies report a dose–response between shift work and cognition, with increasing cognitive losses among those with longer shiftwork tenure.15 ,16 In addition, it may take several years to return to regular cognitive functioning after leaving non-standard schedules.16

    While it is generally agreed that sleep mediates the relationship between shift work and cognitive function,8 ,17 little is known about the influence of other factors which may directly or indirectly affect this relationship. For example, working non-standard hours has also been associated with increased psychosocial and behavioural stressors,18 ,19 triggering neuroendocrine responses. Without sufficient recovery, this may lead to chronic activation of neuroendocrine systems and affect regions in the brain responsible for cognitive performance (ie, prefrontal cortex and hippocampus).20 Stress has also been linked to disturbed sleep and impaired awakening.21 Thus, stress related to shift work may be an important factor affecting cognitive function indirectly through sleep disruptions. Individual-level factors such as age and health may also add further complexity to the relationship. Increasing age has been associated with greater sleep disturbances and changes in circadian rhythms.22 ,23 As such, it has been recognised that older workers may have greater difficulty in adjusting to changes in sleep timing related to shift work.24 However, the effects of sleep deprivation on cognitive function, in terms of memory and reaction speed, may be less sensitive with increasing age.10 Sleep and cognitive impairment is associated with poor health through changes in biological functions and systemic inflammation, or as a side effect of medication taken for illnesses.25 Gaining a better understanding of how these different factors may be potentially associated with each other and their contribution to the relationship between shift work and cognitive function, may help identify areas where occupational health and safety measures could have the most benefit in injury prevention.

    Gender differences in work injury risk among shift workers have also been recently explored.1 ,26 It has been speculated that the higher risk of work injury among women working non-standard shifts26 could be attributed to several factors relating to sleep and cognitive function. Women report more sleep impairments such as inadequate duration and poor quality,27–30 and may be up to 40% more likely to suffer from insomnia than men.31 Women also report spending more time than men on household duties32 ,33 which may leave less time for restorative sleep, a greater degree of work-life conflict34 ,35 and consequently affect fatigue and cognition. Differences in individual-level factors such as health and stress also may ultimately contribute to gender differences in cognition. Women are more likely to report poor health,36 whereas job stress has a greater impact on psychological distress for men.37 Examining differences in factors contributing to sleep and cognition between male and female shift workers may help explain gender differences observed in work injury risk among this group.

    The purpose of this study is to (1) identify the extent to which work and non-work factors mediate the relationship between shift work, sleep and subsequent cognitive function; and (2) determine if the mediating pathways differ for men and women. We hypothesise that work stress will mediate the relationship between shift work and sleep; and that age and health are confounders between sleep and cognitive function. We also hypothesise that the relationship between stress and sleep will be stronger among men, and the relationship between shift schedule and sleep quality will be stronger among women.

    Methods and materials

    Study population

    This study used cross-sectional data from the 2010 National Population Health Survey (NPHS; N=14 633). The NPHS is a Statistics Canada-administered survey of a representative sample of the national population, excluding those living on Indian Reserves and Crown Lands, residents of health institutions, full-time members of the Canadian Forces Bases and some remote areas. We restricted our study population to those between the ages of 15 and 65, who worked in a job in the 12 months prior to being interviewed and whose main source of income came from wages or salaries. Working students were excluded as their main activity was not working for a living.

    Outcome: self-rated cognition problems

    Self-rated cognitive function in the NPHS is one attribute used to derive the Health Utility Index (HUI) measure, a measure of health-related quality of life based on the Comprehensive Health Status Measurement System.38 Cognitive items included respondents' self-rated functioning in memory (How would you describe your usual ability to remember things?) and thinking (How would you describe your usual ability to think and solve day-to-day problems?). Subjective cognition was treated as an ordinal variable with categories of: ‘no cognition problem’, ‘a little difficulty’, ‘somewhat forgetful’, ‘somewhat forgetful/a little difficulty thinking’ and ‘very forgetful/great deal of difficulty thinking’. Self-reported cognition items in the HUI have been shown to have high correlation with the Rey Auditory Verbal Learning Test and other neuropsychological memory tests in adults.39 This measure has been shown to have good reliability (κ=0.70) and construct validity.40

    Primary independent variable: shift type

    Respondents were asked to select one of the eight shift-type categories which described the hours they typically worked in their current job. These categories were collapsed into four shift types as in prior studies1 regular day time (reference group); regular night/graveyard; rotating and other (eg, split shift, on-call, irregular). Shift work was treated as a nominal categorical variable in our analyses.

    Potential mediating variables between shift work and cognition: sleep quality and quantity

    While self-rated sleep quality and estimates of sleep duration are commonly used in sleep diaries and epidemiologic studies,41 there is no consistent measure. However, the self-reported sleep measures used in this study are commonly used questions in sleep diaries42 and across numerous Statistics Canada surveys such as the Canadian Community Health Survey and the General Social Survey to report the prevalence of insomnia in the Canadian population.43

    Sleep behaviour was captured in three questions measuring duration of sleep, frequency of trouble sleeping and frequency of refreshing sleep. We treated sleep as two separate variables representing duration and quality, as prior studies have shown that there is modest or no overlap between constructs.44 Usual sleep duration was ascertained with the question ‘How long do you usually spend sleeping each night?’ and does not include time spent resting. In the NPHS, responses were categorised in 1 hour increments starting at ‘under 2 hours’ to ‘12 hours or more’. To obtain sufficient sample size in each category for analyses, we combined responses for <4 hours, and those which were >9 hours, and was treated as a linear variable.

    Sleep quality was defined as a combination of two dimensions of general frequency of trouble sleeping (How often do you have trouble going to sleep or staying asleep?) and refreshing sleep (How often do you find your sleep refreshing?). Respondents reported the frequency of each on a five-point Likert scale, with 1 representing ‘none of the time’ and 5 representing ‘all of the time’. Responses for refreshing sleep were reverse coded and summed with scores from trouble sleeping to create a composite score of sleep quality (range 2–10). Sleep quality was treated as an ordered categorical variable in our analysis, with higher scores representing worsening conditions.

    Covariates

    We included a number of covariates as potential confounders and mediators in our analyses.

    Work stress was treated as a potential mediator on the relationship between shift work and sleep. In the NPHS, work stress index is based on Karasek's Job Strain model and is a composite score derived from self-ratings of 12 work-related stressors. Higher scores represent higher self-perceived stress.

    Age and health were included as confounders between sleep and subjective cognition, as prior evidence has shown that sleep and cognitive impairments are associated with older ages and poorer health.10 Age was treated as a continuous variable. Respondents in the NPHS self-rated their health from five categories ranging from ‘excellent’ to ‘poor’ with the general health component of the Short Form (SF)-36. We scored responses on a five-point scale using the weights suggested in SF-36 guidelines and treated the measure as a continuous variable in analysis.

    Our hypothesised theoretical model is shown in figure 1.

    Figure 1
    Figure 1

    Hypothesised theoretical pathway for predicting the association of subjective cognitive function with shift work and sleep quality and duration, including mediating and confounding effects of age, health and work stress. The model was fitted to self-reported data from the National Population Health Survey (2010). The sample population was restricted to employed workers between the ages of 15 and 65. Shift work was categorised as day (reference category), night, rotating and other shift types. Sleep quality values ranged from 2 to 10 and was a composite score of frequency of trouble sleeping and reverse-coded frequency of refreshing sleep. Sleep duration was defined as usual time sleeping each night and did not include time spent resting. Values were in increments of 1 hour and ranged from ‘<4 hours’ to ‘more than 9 hours’. Subjective cognitive function included self-rated categories of ‘no cognition problem’, ‘a little difficulty’, ‘somewhat forgetful’, ‘somewhat forgetful/a little difficulty thinking’ and ‘very forgetful/great deal of difficulty thinking’. The reference category for ‘sex’ was female. Age was represented in years. Work stress and health were continuous variables based on respondents' self-reported ratings. *Sleep quality and quantity were treated as separate variables. Sleep quality was a composite score of frequency of trouble sleeping and reverse-coded frequency of refreshing sleep.

    Statistical analyses and model building

    Path modelling was used to gain a better understanding of the complex direct and indirect relationships between shift work, stress, sleep and subjective cognitive function.45 The benefit of path modelling over traditional regression methods is that it can examine direct and indirect pathways between distal (eg, shift work) and proximal predictors (eg, sleep, age, health, stress) with the outcome measure (cognitive function).46 By relaxing the assumption of independence required with logistic regression models, this method also allows collinearity between the independent variables to be examined.47

    A path model was outlined to estimate the direct, indirect and total effects of all predictive variables on cognitive function. Direct effects represent the association between the predictor and the outcome that is not mediated through other variables, adjusted for all potential confounders of the predictor–outcome relationship. Indirect effects refer to effects between the predictor and the outcome that are mediated by one or more intermediate variables. Total effects are the sum of the parameter estimates for direct and indirect effects.47 Analyses were conducted with MPlus V.7.3 (Muthén and Muthén, Los Angeles, California, USA), using a robust weighted least squares estimator. p Values <0.05 (two-sided) were considered statistically significant. To account for the probability of selection and non-response to the original survey, all results were weighted using sampling weights provided by Statistics Canada.

    All hypothesised pathways in our theoretical model were included in the initial analyses. As part of the model building strategy, pathways where no relationship was observed were removed from further analyses. Additional correlations between errors of predictor variables were added where applicable. To assess our model fit, we compared several indicators commonly used in path analyses: the root mean square error of approximation (RMSEA), Comparative Fit Index (CFI) and Tucker Lewis Index (TLI). We considered a good model fit where RMSEA upper bound <0.08, CFI>0.90 and TLI>0.90.48

    We used a four-step process to explore if the mediating pathway between shift work and cognitive function differed between men and women. In the initial step, we developed a preliminary model based on theoretical assumptions which achieved good model fit. We then examined this model separately for men and women for model fit. An unconstrained model was developed by combining the two gender-specific models and all parameters were allowed to be estimated independently for men and women. To test for significantly different pathways between genders, we conducted multigroup analyses using the grouping command in MPlus. We constrained each pathway one at a time, such that the estimates for men and women would be forced to be equal, and compared results with the unconstrained model using a χ2 test for nested models. In the final model, paths that were significantly different between genders would be estimated freely, thus allowing separate parameter estimates for men and women.

    All descriptive analyses were completed with Stata V.12 (StataCorp LP, College Station, Texas, USA), using procedures for complex survey data.

    Results

    Study population

    In 2010, 14 663 respondents participated in the NPHS. We excluded responses from those who reported not working as an employee in the 12 months prior to be interviewed or who were not 16–65 years old (N=9724), and those who were currently students (N=684). As such, our final study population consisted of 4255 respondents.

    Table 1 presents descriptive information on the study population. Slightly more women than men were included in our sample (54% women). The average age of all workers was 42.5 (SE=0.27). There was no difference between men and women in subjective cognitive function. However, women reported fewer hours of sleep (p=0.015), more trouble sleeping (p<0.001), lower frequency of refreshing sleep (p=0.001) and more work-related stress (p<0.001). Our distribution of workers by shift types (ie, daytime only 75.4%, regular night 5.2%, rotating 9.7%, other 9.7%) is similar to estimates among the Canadian workforce population reported in prior studies.1

    View this table:
    Table 1

    Descriptive characteristics of study population

    Model building

    Results of the unconstrained model, stratified by gender, are presented in online supplementary table S1. In the preliminary stages of testing our theoretical model, we found that sleep duration was not significantly associated with subjective cognitive function and was therefore removed from the model. Preliminary model fit after the removal of these two variables was good (RMSEA upper bound=0.023, CFI=0.983 and TLI=0.958).

    Supplementary table

    Unconstrained Model, Gender Stratified

    In our gender analyses, we found that for women, variables directly associated with subjective cognition included sleep quality and health, whereas for men, age was significantly associated. For both genders, health and work stress were significantly associated with sleep quality, but night shifts directly impacted sleep quality for women only. However, differences in estimates for men and women for all pathways were not statistically different (see online supplementary table S1). As such, our final model included gender as a confounder between sleep and work stress (figure 2). This model also had acceptable model fit as assessed by multiple fit indices (RMSEA upper bound=0.032, CFI=0.952, TLI=0.889).

    Figure 2
    Figure 2

    Results from path analyses for predicting the association of subjective cognitive function with shift work and sleep quality, including mediating and confounding effects of age, health and work stress. The model was fitted to self-reported data from the National Population Health Survey (2010). The sample population was restricted to employed workers between the ages of 15 and 65. Shift work was categorised as day (reference category), night, rotating and other shift types. Sleep quality values ranged from 2 to 10 and was a composite score of frequency of trouble sleeping and reverse-coded frequency of refreshing sleep. Subjective cognitive function included self-rated categories of ‘no cognition problem’, ‘a little difficulty’, ‘somewhat forgetful’, ‘somewhat forgetful/a little difficulty thinking’ and ‘very forgetful/great deal of difficulty thinking’. The reference category for ‘sex’ was female. Age was represented in years. Work stress and health were continuous variables based on respondents' self-reported ratings. Pathways marked with an asterisk (*) are significant (p<0.05).

    Direct and indirect effects of shift work on sleep quality and cognitive function

    Table 2 presents the parameter estimates for direct, indirect and total effects for shift schedule on sleep quality. Figure 2 provides a graphical representation of pathways in our final model. All categories of shift work (ie, night, rotating and other) were significantly and indirectly associated with poorer sleep quality through increasing work stress (p <0.001). However, total effects were significant only for night shifts (p=0.006). We also found that being female, poorer self-reported health and higher levels of work stress were significant direct effects on sleep quality. In addition, there was an indirect pathway between women and poor sleep quality via increased work stress (p≤0.001).

    View this table:
    Table 2

    Direct, indirect and total effects on sleep quality*

    Direct, indirect and total effects for study variables on subjective cognitive function are reported in table 3 and represented in figure 2. Direct effects of poor sleep quality on lower subjective cognitive function were found (p<0.001) but not for categories of shift work (p>0.05). However, the relationship between night shifts to poorer subjective cognitive function were significantly mediated through the pathway linking work stress to sleep quality (p=0.011), although total effects were not significant (p=0.121). In addition, younger age and poorer health were associated with both sleep quality and subjective cognitive function. The relationship between sleep quality and subjective cognitive function remained significant after taking these associations into account.

    View this table:
    Table 3

    Direct, indirect and total effects on subjective cognition

    Discussion

    Previous studies have theorised that shift work results in poor sleep quality, which in turn affects cognitive function. This study is among the first to test these relationships using path modelling aimed at better understanding the direct and indirect effects among shift work, sleep quality and subjective cognition along with gender, work stress, age and health. The findings highlight that sleep quality directly impacts cognition.49 However, our finding that sleep duration does not impact cognition contradicts findings from prior studies.14 The reasons for this are unclear. Sleep quality and quantity are separate constructs44 and it may be possible that while a respondent obtains a sufficient amount of sleep, it may not be good quality or restful sleep, thus affecting subjective cognition. Our outcome measure of subjective cognition also differs from objective measures used in prior studies, which may have resulted in differences in study findings. We also observed that greater work stress is a significant mediator in the relationship between shift work and sleep quality. Sex, age and health also played important roles in terms of understanding differences in stress and sleep quality among women and men, and in the contribution of age and poorer health in understanding the relationship between sleep quality and subjective cognition. As such, this study adds more complexity and nuanced findings in our understanding the impact of shift work and sleep quality in the lives of workers.

    Prior studies have reported that shift work is associated with greater work stress.50 However, little is known about what specific components of non-standard schedules may be contributing to this significant relationship. It may be that the pace or intensity of work-related demands may affect job stress and ultimately sleep. It has also been suggested that the lack of control over work hours associated with shift work and especially variable shifts may result in higher work stress.18 There may also be other elements of stress that were not assessed by the measure included in this research and that would help explain the mediating role of work stress between shift work and sleep quality.51 The findings of this study suggest that the components of work stress deserve additional attention in future research. They also suggest that the relationship between shift work and sleep quality found in previous studies may be understood in terms of increases in work stress associated with non-regular shifts.

    Additional research examining sex and gender differences would also be valuable in understanding the role of work stress in mediating shift work and sleep quality. Women in the study reported more sleep quality difficulties and greater work stress. Previous research finds that women often report having less job control than men52 and may take on additional roles for caregiving and household activities.32 These factors in addition to shift work may have combined to result in greater work stress and poorer sleep quality among women. At the same time, it is important to highlight that the model itself, with shiftwork being mediated by work stress and this, in turn, relating to sleep quality and subjective cognition was similar for women and men. Differences were in the level of stress and sleep quality problems, not in the nature or types of the relationships experienced by either sex.

    The inclusion of age and health also helped illuminate the association between sleep quality and subjective cognition. Not surprisingly, older adults and those with poorer health reported worse sleep quality. Moreover, poorer sleep quality and health were associated with poorer subjective cognition, which is consistent with other findings.25 Of interest, however, was that younger workers in the study reported poorer subjective cognition than their middle-aged and older-aged working counterparts. Reasons for this are not clear. In keeping with population-level data, the measure of subjective cognition was fairly simplistic and emphasised memory. In addition, given that all participants were employed and 65 years of age or less, the likelihood of substantial self-reported cognitive problems was reduced in this sample. It may be that the younger workers in the study had a different standard or interpretation of the cognitive questions. Cognition like memory can also be influenced by a range of factors, including other activities, which may have differed by age. Additional research with more detailed measures of cognition that are validated across age groups is needed, as well as research that takes into account a range of specific job activities and other lifestyle factors that might differ by age.

    There are several limitations that need to be acknowledged in this study. The data used in this study did not include a measure of shift work experience. As such, it is unclear if shift work tenure may have influenced the mediating role of work stress in the relationship between shift work and sleep quality. It may be possible that extended shift work tenure results in poorer quality sleep due to chronic work stress, or it may demonstrate a healthy worker effect as those who have remained in shift work the longest have been able to opt for less stressful work or have adapted to the occupational demands associated with shift work. Future studies are needed to examine this further. Our data also lacked information about circadian type, which has been shown to be significantly associated with a variety of cognitive functions including learning, arousal and information processing.53 While it has been suggested that cognitive performances may differ between morning type and evening types through the 24-hour clock,53 there is a paucity of research which examines this. As such, future investigations examining the effects of circadian type on cognition over the course of the day may help inform development of job task scheduling and rotation across different shift types to ensure safer work practices. Although some validation work on the subjective cognition measure has occurred comparing the measure with experimenter-administered cognitive tasks39 ,40 and many of the predicted findings were confirmed, the use of self-reported measures like cognitive functioning may be biased due to recall error and social desirability. For example, some respondents may not want to report cognitive difficulties, especially as they age, or may not be aware of small declines.39 Cognition is also a complex concept, which in addition to memory and problem solving, includes other aspects such as motor skills making it a difficult outcome to accurately quantify. We also suggest caution in the interpretation of the mediating role of sleep quality between shift work and subjective cognition. Although the indirect pathway between shift work and subjective cognitive function was statistically significant, the total effect was not, even though these estimates were both in the same direction. This may have resulted because the detection of significant indirect effects requires less study power and stronger assumptions about unmeasured confounding compared with total or direct effects.54 Finally, the cross-sectional study design limits our ability examine causal relationships. One of the strengths of the path model approach is the need to present an a priori theoretical causal model of the relationships between study variables. Fit statistics then indicate the extent to which this theoretical model is supported by the data. However, it is possible, in particular with cross-sectional data, that the direction of the arrows between study variables may be in the opposite direction to those specified in our theoretical model (see figure 1). For example, self-rated health might be an outcome of poor sleep quality/duration or lower cognition, rather than a predictor as specified in our models. With cross-sectional data, it is impossible to disentangle these relationships. If self-rated health were an outcome, rather than a predictor, including it in our model will have reduced the observed effects of variables predicting sleep quality and cognition, as such the estimates presented in this paper may be underestimated. In short, longitudinal studies are needed to validate our findings. One possible method would be to use multiwave cross-lagged path analyses to examine the effect of shift work on sleep and cognition at baseline and in subsequent years. Self-rated health and stress would also be included in baseline and follow-up years to examine their impact as possible predictors in current years and outcome measures in subsequent years. An example of this method has been used in prior longitudinal path analyses examining work stress and health as predictors and outcomes across time.55

    Despite the limitations, the use of a national population survey allows for generalisation to the larger workforce population. Our use of path modelling, which has been underused in occupational epidemiology,56 allowed us to expand on existing studies of shift work, sleep and subjective cognition by examining the direct and indirect factors contributing to the mediating pathway. Our study findings suggest that the relationships among shift work, sleep quality and subjective cognitive functioning are complex; and that other variables such as work stress, age, sex and health contribute important information about sleep quality and subjective cognitive functioning. These results point to the need for occupational health and safety efforts to reduce the effects of sleep disruption and subjective cognitive impairment among non-standard shift workers while considering a range of factors, like work stress and health, as well as greater attention to potential age differences.

    References

    1. ↵
      1. Wong IS,
      2. McLeod CB,
      3. Demers PA
      . Shift work trends and risk of work injury among Canadian workers. Scand J Work Environ Health 2011;37:54–61. doi:10.5271/sjweh.3124
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      1. Mustard C,
      2. Chambers A,
      3. McLeod C, et al
      . Work injury risk by time of day in two population-based data sources. Occup Environ Med 2013;70:49–56. doi:10.1136/oemed-2012-100920
      OpenUrlAbstract/FREE Full Text
    3. ↵
      1. Stimpfel AW,
      2. Brewer CS,
      3. Kovner CT
      . Scheduling and shift work characteristics associated with risk for occupational injury in newly licensed registered nurses: an observational study. Int J Nurs Stud 2015;52:1686–93. doi:10.1016/j.ijnurstu.2015.06.011
      OpenUrlCrossRefPubMed
    4. ↵
      1. Kling RN,
      2. McLeod CB,
      3. Koehoorn M
      . Sleep problems and workplace injuries in Canada. Sleep 2010;33:611–18.
      OpenUrlPubMedWeb of Science
    5. ↵
      1. Hartenbaum N,
      2. Zee P
      . Shift work and sleep: optimizing health, safety, and performance. J Occup Environ Med 2011;53(5 Suppl):S1–10.
      OpenUrlPubMed
    6. ↵
      1. Geiger-Brown J,
      2. Rogers VE,
      3. Trinkoff AM, et al
      . Sleep, sleepiness, fatigue, and performance of 12-hour-shift nurses. Chronobiol Int 2012;29:211–19. doi:10.3109/07420528.2011.645752
      OpenUrlCrossRefPubMed
    7. ↵
      IARC Working Group on the Evaluation of Carcinogenic Risks to Humans. Painting, firefighting, and shiftwork. IARC Monogr Eval Carcinog Risks Hum 2010;98:9–764.
      OpenUrlPubMed
    8. ↵
      1. Akerstedt T,
      2. Wright K
      . Sleep loss and fatigue in shift work and shift work disorder. Sleep Med Clin 2009;4:257–71. doi:10.1016/j.jsmc.2009.03.001
      OpenUrlCrossRefPubMed
    9. ↵
      1. Wright KP,
      2. Lowry CA,
      3. Lebourgeois MK
      . Circadian and wakefulness-sleep modulation of cognition in humans. Front Mol Neurosci 2012;5:50. doi:10.3389/fnmol.2012.00050
      OpenUrlCrossRefPubMed
    10. ↵
      1. Killgore WDS
      . Effects of sleep deprivation on cognition. Prog Brain Res 2010;185:105–29. doi:10.1016/B978-0-444-53702-7.00007-5
      OpenUrlCrossRefPubMedWeb of Science
    11. ↵
      1. MacHi MS,
      2. Staum M,
      3. Callaway CW, et al
      . The relationship between shift work, sleep, and cognition in career emergency physicians. Acad Emerg Med 2012;19:85–91. doi:10.1111/j.1553-2712.2011.01254.x
      OpenUrlCrossRefPubMed
    12. ↵
      1. Ansiau D,
      2. Wild P,
      3. Niezborala M, et al
      . Effects of working conditions and sleep of the previous day on cognitive performance. Appl Ergon 2008;39:99–106. doi:10.1016/j.apergo.2007.01.004
      OpenUrlCrossRefPubMed
    13. ↵
      1. Tadinac M,
      2. Sekulić A,
      3. Hromatko I, et al
      . Age and individual sleep characteristics affect cognitive performance in anesthesiology residents after a 24-hour shift. Acta Clin Croat 2014;53:22–30.
      OpenUrlPubMed
    14. ↵
      1. Van Dongen HP,
      2. Maislin G,
      3. Mullington JM, et al
      . The cumulative cost of additional wakefulness: dose-response effects on neurobehavioural functions and sleep physiology from chronic sleep restriction and total sleep restriction. Sleep 2003;26:117–26.
      OpenUrlPubMedWeb of Science
    15. ↵
      1. Rouch I,
      2. Wild P,
      3. Ansiau D, et al
      . Shiftwork experience, age and cognitive performance. Ergonomics 2005;48:1282–93. doi:10.1080/00140130500241670
      OpenUrlCrossRefPubMedWeb of Science
    16. ↵
      1. Marquie J-C,
      2. Tucker P,
      3. Folkard S, et al
      . Chronic effects of shift work on cognition: findings from the VISAT longitudinal study. Occup Environ Med 2015;72:258–64. doi:10.1136/oemed-2013-101993
      OpenUrlAbstract/FREE Full Text
    17. ↵
      1. Gold DR,
      2. Rogacz S,
      3. Bock N, et al
      . Rotating shift work, sleep, and accidents related to sleepiness in hospital nurses. Am J Public Health 1992;82:1011–14. doi:10.2105/AJPH.82.7.1011
      OpenUrlCrossRefPubMedWeb of Science
    18. ↵
      1. Puttonen S,
      2. Härmä M,
      3. Hublin C
      . Shift work and cardiovascular disease—pathways from circadian stress to morbidity. Scand J Work Environ Health 2010;36:96–108. doi:10.5271/sjweh.2894
      OpenUrlCrossRefPubMedWeb of Science
    19. ↵
      1. Boggild H,
      2. Knutsson A
      . Shift work, risk factors and cardiovascular disease. Scand J Work Environ Health 1999;25:85–99. doi:10.5271/sjweh.410
      OpenUrlCrossRefPubMedWeb of Science
    20. ↵
      1. Luine V
      . Recognition memory tasks in neuroendocrine research. Behav Brain Res 2015;285:158–64. doi:10.1016/j.bbr.2014.04.032
      OpenUrlCrossRefPubMed
    21. ↵
      1. Akerstedt T,
      2. Knutsson A,
      3. Westerholm P, et al
      . Sleep disturbances, work stress and work hours: a cross-sectional study. J Psychosom Res 2002;53:741–8. doi:10.1016/S0022-3999(02)00333-1
      OpenUrlCrossRefPubMedWeb of Science
    22. ↵
      1. Monk TH
      . Aging human circadian rhythms: conventional wisdom may not always be right. J Biol Rhythms 2005;20:366–74. doi:10.1177/0748730405277378
      OpenUrlAbstract/FREE Full Text
    23. ↵
      1. Redline S,
      2. Kirchner HL,
      3. Quan SF, et al
      . The effects of age, sex, ethnicity, and sleep-disordered breathing on sleep architecture. Arch Intern Med 2004;164:406–18. doi:10.1001/archinte.164.4.406
      OpenUrlCrossRefPubMedWeb of Science
    24. ↵
      1. Akerstedt T,
      2. Torsvall L
      . Shift work. Shift-dependent well-being and individual differences. Ergonomics 1981;24:265–73.
      OpenUrlCrossRefPubMedWeb of Science
    25. ↵
      1. Alwin DF,
      2. Hofer SM
      . Health and cognition in aging research. J Gerontol B Psychol Sci Soc Sci 2011;66(Suppl 1):i9–16. doi:10.1093/geronb/gbr051
      OpenUrlPubMed
    26. ↵
      1. Wong I,
      2. Smith P,
      3. Mustard C, et al
      . For better or worse? The association of work injury risk with changing shift work schedules. Scand J Work Environ Health 2014;40:621–30.
      OpenUrlCrossRefPubMed
    27. ↵
      1. Bixler EO,
      2. Vgontzas AN,
      3. Lin HM, et al
      . Insomnia in Central Pennsylvania. J Psychosom Res 2002;53:589–92. doi:10.1016/S0022-3999(02)00450-6
      OpenUrlCrossRefPubMedWeb of Science
    28. ↵
      1. Hohagen F,
      2. Rink K,
      3. Schramm E, et al
      . Prevalence and treatment of insomnia in general practice: a longitudinal study. Eur Arch Psychiatry Clin Neurosci 1993;242:329–36. doi:10.1007/BF02190245
      OpenUrlCrossRefPubMedWeb of Science
    29. ↵
      1. Ohayon M
      . Epidemiological study on insomnia in a general population. Sleep 1996;19:7–15.
      OpenUrl
    30. ↵
      1. Partinen M,
      2. Hublin C
      . Principles and practice of sleep medicine. 3rd edn. 2000:558–79.
    31. ↵
      1. Zhang B,
      2. Wing Y-K
      . Sex differences in insomnia: a meta-analysis. Sleep 2006;29:85–93.
      OpenUrlPubMedWeb of Science
    32. ↵
      1. Berntsson L,
      2. Lundberg U,
      3. Krantz G
      . Gender differences in work-home interplay and symptom perception among Swedish white-collar employees. J Epidemiol Community Health 2006;60:1070–6. doi:10.1136/jech.2005.042192
      OpenUrlAbstract/FREE Full Text
    33. ↵
      1. Higgins C,
      2. Duxbury L,
      3. Lee C
      . Balancing work and family: a study of the Canadian private sector employees. London: National Centre for Management Research and Development, Western Business School, the University of Western Ontario, 1992.
    34. ↵
      1. Gignac M,
      2. Backman C,
      3. Kaptein S, et al
      . Tension at the borders: perceptions of role overload, conflict, strain and facilitation in work, family and health roles among employed individuals with arthritis. Rheumatology 2012;51:324–32. doi:10.1093/rheumatology/ker317
      OpenUrlAbstract/FREE Full Text
    35. ↵
      1. Hammig O,
      2. Bauer G
      . Work-life imbalance and mental health among male and female employees in Switzerland. Int J Public Health 2009;54:88–95. doi:10.1007/s00038-009-8031-7
      OpenUrlCrossRefPubMedWeb of Science
    36. ↵
      1. Oksuzyan A,
      2. Crimmins E,
      3. Saito Y, et al
      . Cross-national comparison of sex differences in health and mortality in Denmark, Japan and the US. Eur J Epidemiol 2010;25:471–80. doi:10.1007/s10654-010-9460-6
      OpenUrlCrossRefPubMed
    37. ↵
      1. Vermeulen M,
      2. Mustard C
      . Gender differences in job strain, social support at work, and psychological distress. J Occup Health Psychol 2000;5:428–40. doi:10.1037/1076-8998.5.4.428
      OpenUrlCrossRefPubMed
    38. ↵
      1. Horsman J,
      2. Furlong W,
      3. Feeny D, et al
      . The Health Utilities Index (HUI): concepts, measurement properties and applications. Health Qual Life Outcomes 2003;1:54. doi:10.1186/1477-7525-1-54
      OpenUrlCrossRefPubMed
    39. ↵
      1. Findlay L,
      2. Bernier J,
      3. Tuokko H, et al
      . Validation of cognitive functioning categories in the Canadian Community Health Survey. Health Rep 2010;21:2452–66.
      OpenUrl
    40. ↵
      1. Mulhern R
      . Correlation of the Health Utilities Index Mark 2 cognition scale and neuropsychological functioning among survivors of childhood medulloblastoma. Int J Cancer Suppl 1999;12:91–4. doi:10.1002/(SICI)1097-0215(1999)83:12+<91::AID-IJC17>3.0.CO;2-O
      OpenUrlPubMed
    41. ↵
      1. Ohayon M,
      2. Lemoine P,
      3. Arnaud-Briant V, et al
      . Prevalence and consequences of sleep disorders in a shift worker population. J Psychosom Res 2002;53: 577–83. doi:10.1016/S0022-3999(02)00438-5
      OpenUrlCrossRefPubMedWeb of Science
    42. ↵
      1. Ong J,
      2. Arnedt T,
      3. Gehrman P
      . Insomnia diagnosis, assessment and evaluation. In: Kryger M, Roth T, Dement W, eds. Principles and practice of sleep medicine edn. Philadelphia, PA, et al Elsevier, 2017:785–93.
    43. ↵
      1. Tjepkema M
      . Insomnia. Health Rep 2005;17:9–25.
      OpenUrlPubMed
    44. ↵
      1. Short MA,
      2. Gradisar M,
      3. Lack LC, et al
      . The impact of sleep on adolescent depressed mood, alertness and academic performance. J Adolesc 2013;36:1025–33. doi:10.1016/j.adolescence.2013.08.007
      OpenUrlCrossRefPubMed
    45. ↵
      1. Arlinghaus A,
      2. Lombardi DA,
      3. Willetts JL, et al
      . A structural equation modeling approach to fatigue-related risk factors for occupational injury. Am J Epidemiol 2012;176:597–607. doi:10.1093/aje/kws219
      OpenUrlAbstract/FREE Full Text
    46. ↵
      1. Weitkunat R,
      2. Wildner M
      . Exploratory causal modeling in epidemiology: are all factors created equal? J Clin Epidemiol 2002;55:436–44.
      OpenUrlCrossRefPubMedWeb of Science
    47. ↵
      1. Kline R
      . Principles and practice of structural equation modeling. 3rd edn. New York, USA: Guilford Press, 2010.
    48. ↵
      1. Hu LT,
      2. Bentler PM
      . Evaluating model fit. In: Hoyle R, ed. Equation modeling: concepts, issues, and applications. Thousand Oaks, CA: Sage Publications, Inc., 1995:76–99.
    49. ↵
      1. Miller MA,
      2. Wright H,
      3. Ji C, et al
      . Cross-sectional study of sleep quantity and quality and amnestic and non-amnestic cognitive function in an ageing population: The English Longitudinal Study of Ageing (ELSA). PLoS ONE 2014;9:e100991. doi:10.1371/journal.pone.0100991
      OpenUrlCrossRefPubMed
    50. ↵
      1. Lin P,
      2. Chen C,
      3. Pan S, et al
      . The association between rotating shift work and increased occupational stress in nurses. J Occup Health 2015;57:307–15. doi:10.1539/joh.13-0284-OA
      OpenUrlCrossRefPubMed
    51. ↵
      1. Ala-Mursula L,
      2. Vahtera J,
      3. Kivimäki M, et al
      . Employee control over working times: associations with subjective health and sickness absences. J Epidemiol Community Health 2002;56:272–8.
      OpenUrlAbstract/FREE Full Text
    52. ↵
      1. LaMontagne AD,
      2. Krnjacki L,
      3. Kavanagh AM, et al
      . Psychosocial working conditions in a representative sample of working Australians 2001-2008: an analysis of changes in inequalities over time. Occup Environ Med 2013;70:639–47. doi:10.1136/oemed-2012-101171
      OpenUrlAbstract/FREE Full Text
    53. ↵
      1. Adan A,
      2. Archer SN,
      3. Hidalgo MP, et al
      . Circadian typology: a comprehensive review. Chronobiol Int 2012;29:1153–75. doi:10.3109/07420528.2012.719971
      OpenUrlCrossRefPubMed
    54. ↵
      1. Loeys T,
      2. Moerkerke B,
      3. Vansteelandt S
      . A cautionary note on the power of the test for the indirect effect in mediation analysis. Front Psychol 2015;5:1–8. doi:10.3389/fpsyg.2014.01549
      OpenUrl
    55. ↵
      1. Ibrahim S,
      2. Smith P,
      3. Muntaner C
      . A multi-group cross-lagged analyses of work stressors and health using Canadian National sample. Soc Sci Med 2009; 68:49–59. doi:10.1016/j.socscimed.2008.10.019
      OpenUrlCrossRefPubMed
    56. ↵
      1. Davis ME
      . Structural equation models in occupational health: an application to exposure modelling. Occup Environ Med 2012;69:184–90. doi:10.1136/oem.2010.063032
      OpenUrlAbstract/FREE Full Text

    Footnotes

    • Contributors ISW conceptualised the study idea, developed the study design, and performed data analyses and interpretation. PMS and MAMG contributed to study design, data analyses and interpretation. SI contributed to data analyses and interpretation. CAM contributed to data acquisition and data interpretation. All authors contributed to the editorial preparation of this paper.

    • Funding Funding for ISW was provided by the Mustard Post-doctoral Fellowship from the Institute for Work & Health (IWH), and the Oak Ridge Institute for Science and Education (ORISE) through an interagency agreement with the National Institute for Occupational Safety and Health (NIOSH). PMS is supported by a Chair in Gender, Work & Health from the Canadian Institutes of Health Research. IWH operates with the support of the Province of Ontario.

    • Disclaimer The views expressed in this document are those of the authors and do not necessarily reflect those of the Province of Ontario or ORISE.

    • Competing interests None declared.

    • Ethics approval Ethical approval for this study was obtained through the University of Toronto, Health Sciences' Ethics Committee.

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