Article Text

Download PDFPDF

Association between psychosocial work conditions and latent alcohol consumption trajectories among men and women over a 16-year period in a national Canadian sample
  1. Kathleen G Dobson1,2,
  2. Selahadin Ibrahim1,
  3. Mahée Gilbert-Ouimet1,3,
  4. Cameron A Mustard1,2,
  5. Peter M Smith1,2,4
  1. 1 Institute for Work and Health, Toronto, Ontario, Canada
  2. 2 Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada
  3. 3 Centre de Recherche du CHU de Québec, Quebec City, Quebec, Canada
  4. 4 Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, Australia
  1. Correspondence to Kathleen G Dobson, Institute for Work & Health, Toronto, Ontario M5G2E9, Canada; kathleen.dobson{at}


Background It is unclear how psychosocial working conditions influence future alcohol consumption. Using group-based trajectory modelling, this study aimed to determine: the number of latent alcohol consumption trajectories over 16 years in a representative sample of the Canadian workforce; the association between psychosocial working conditions and longitudinal alcohol consumption; and if the association between psychosocial work factors and longitudinal alcohol consumption differed among men and women.

Methods We included 5458 employed adults from the longitudinal Canadian National Population Health Survey. Average daily alcohol consumption was measured every 2 years from 1994 to 2010. Psychosocial work factors were measured in 1994 using the Job Content Questionnaire. Group-based trajectory modelling was used to derive the appropriate number of alcohol behaviour trajectories. The association between psychosocial work factors and alcohol trajectory membership was estimated using multinomial logistic regression. Models were stratified by sex to determine if these associations differed among men and women.

Results Three alcohol consumption trajectories were present: non-drinkers, light drinkers (0.5–1 drinks/day) and moderate drinkers (2–3 drinks/day). Higher workplace physical exertion and lower social support levels were associated with membership in the moderate drinking trajectory. Among men, lower psychological demands and higher physical exertion levels were associated with membership in the moderate drinking trajectory. Among women, lower levels of physical exertion were associated with membership in the light drinking trajectory, and higher psychological demand levels were associated with membership in the moderate drinking trajectory.

Conclusions Our study suggests that workplace physical exertion and psychological demands may be associated with different alcohol consumption trajectories among men and women.

  • alcohol
  • psychosocial work stress
  • occupational health
  • latent trajectory
  • work environment

Statistics from

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.


Heavy alcohol use—defined as more than 7 standard drinks per week for women and 14 for men1—is associated with over 200 injury and disease conditions, including cancer, mood disorders and cardiovascular disease.2 Heavy alcohol use can significantly affect an individual’s health, family, socioeconomic status and employment. The association between employment factors and alcohol use has been of great interest among researchers, employers and occupational health professionals over the past three decades.3 One of the most studied predictors of heavy alcohol use has been psychosocial work conditions, defined as the opportunities available for an individual to meet their needs related to well-being, productivity, self-experience and positive interactions with others within the workplace.4

Literature is divided about if and what psychosocial working conditions influence alcohol consumption. A recent pooled meta-analysis of 12 cross-sectional studies found that non-drinkers and heavy drinkers reported higher job strain (high psychological demands and low decision authority) than moderate drinkers, while intermediate drinkers reported lower job strain compared with moderate drinkers.5 This meta-analysis also examined four longitudinal studies, suggesting that job strain was not associated with engaging in heavy drinking over time.5 Additionally, other prominent psychosocial work factors including skill discretion, social support and physical exertion have also been studied with mixed findings.6–9 However, the vast majority of literature has been cross-sectional,5 10–13 limiting the ability to infer causation. Of available longitudinal studies, the majority only include one or two aspects of the psychosocial work environment or only one assessment of alcohol use;8 14 15 this fails to account for how these associations are impacted by other psychosocial work factors over one’s career.

Additionally, some studies have either studied male-only samples7 16 or specific occupational groups (eg, military) that are male dominated,17 18 limiting our knowledge of how the psychosocial work environment impacts alcohol use among women. Some literature has explored if the association between psychosocial work conditions and alcohol use differs among men and women.11 19–22 These studies suggest that the association between psychosocial working conditions—such as physical demands at work and high workload—and heavy alcohol use is stronger in men compared with women.13 21 However, given the fact that many alcohol guidelines and recommendations are different for men and women, that men and women may have different experiences in the workplace, and that previous studies include different aspects of the work environment, it is important to better understand if differences exist in the relationships between the psychosocial work environment and alcohol consumption for working men and women.

There has been a call within the literature to develop more accurate longitudinal models of how psychosocial work conditions affect alcohol use trajectories.14 Given the vast heterogeneity of employment types, using a latent trajectory approach to identify the most common drinking behaviours within the workforce may lead to more accurate information about how drinking behaviours are truly affected by psychosocial work conditions among men and women. The present study explores how psychosocial work conditions were associated with alcohol consumption behaviours over a 16-year period by determining: (1) the number of latent alcohol consumption trajectories in a national sample of employed Canadians; (2) if psychosocial work conditions were associated with membership to a particular alcohol use trajectory; and (3) if the association between psychosocial work conditions and alcohol trajectories differed for men compared with women.



This study used data from the longitudinal component of the Canadian National Population Health Survey (NPHS). The NPHS was a national, representative longitudinal study of Canadians that occurred every 2 years from 1994 to 2010 with the goal of identifying the health, demographic and social characteristics of Canadians.23 The NPHS employed a two-stage sampling design of clusters and dwellings and in 1994 surveyed at least 1200 homes in each province, resulting in a sample of 17 276 of all ages.23 The overall sample decreased slightly over time, primarily due to refusal to participate, non-response and participant death.

For this study, the sample was restricted to employed labour force participants in 1994, working at least 15 hours a week, and who were not self-employed (n=6140). The analytic sample included participants with at least two valid responses on alcohol use between 1994 and 2010 and complete information on covariates measured in 1994 (n=5458). Participants excluded from the analytic sample were more likely to be male, older, and drink slightly less in 1994; there were no differences regarding psychosocial working conditions. Roughly 70% of the analytic sample held employment at all cycles over the 16 years.

Outcome: daily alcohol use

Alcohol use was defined by the average number of alcoholic drinks consumed per day. This metric of alcohol use was used as it was measured consistently over all nine cycles. Participants were asked if they had previously consumed alcohol within the past 12 months; participants who reported consuming alcohol were asked how many drinks they consumed each day of the week for the week prior to the interview. The number of alcoholic drinks consumed over the week was summed together and divided by 7 for each participant to obtain the average number of drinks consumed daily.

Alcohol use was examined at each cycle, resulting in nine measurements over 16 years. To improve the convergence of the latent trajectory models, daily alcohol use was truncated to the 99th percentile; therefore, daily alcohol consumption ranged from 0 to 4 or 5 drinks per day over the nine cycles.

Exposures: psychosocial working conditions

The 1994 NPHS included an abbreviated measure of the Karasek and Theorell’s Job Content Questionnaire, containing 12 questions on perceptions of various workplace factors.24 The psychosocial working conditions explored in this study included skill discretion, decision authority, psychological demands, job insecurity, physical exertion and workplace social support (table 1). Respondents answered each question on a 5-point Likert scale (0=strongly agree, 4=strongly disagree). Questions were then combined to make a scale for each dimension. Working conditions were used as continuous variables in analyses. The internal consistency of psychosocial work factors ranges from 0.34 to 0.61,25 with these relatively low Cronbach’s alphas attributed to the abbreviated nature of each scale.

Table 1

Description of psychosocial work factors from the 1994 National Population Health Survey


All covariates were measured in 1994. Demographic covariates included sex (male, female); age (continuous in years); country of birth (Canada, other); race (Caucasian, non-Caucasian); highest level of education (less than high school education, high school diploma, college diploma, completion of a university degree); marital status (married/common law, single/divorced/widowed); dependent status (has children under 12 years, no children under 12 years); province of residence (Ontario, Quebec, Pacific Provinces, Atlantic Provinces); and residence in a rural or urban location (defined as a city with a population concentration of 1000 or more and a population density of 400 or more per square kilometre based on the previous census). Employment covariates included the shift schedule of the main employment (regular, rotating, irregular/on call) and employment status (full, part time). Health restrictions (yes, no), defined as a health condition impacting the individual’s ability to go to work, school or be involved in the community, was also included in the model.

Statistical analysis

Descriptive statistics of alcohol use, psychosocial work factors and covariates were computed. Correlations between all variables assessed over the nine cycles were examined; study covariates did not change heavily over time and thereby were treated as time invariant in analyses.

A group-based trajectory modelling (GBTM) approach using a Poisson distribution to derive latent alcohol consumption behaviours trajectories was used for this study. GBTM is a subset of finite mixture modelling with the purpose of modelling latent trajectories.26 This technique does not make assumptions about the distribution of the population (ie, if each latent class is truly a distinct subpopulation) but does have a conditional independence assumption (ie, an individual’s repeated measures of an outcome over time are independent conditional on membership in a specific trajectory).27

To answer the first objective, an iterative modelling process of exploring different models was completed, specifying different number of latent classes, trajectory growth factors (ie, intercepts and slopes) and constraining and varying the variance of the growth factors. We explored models specifying 1–5 latent classes. When specifying the growth factors, we explored models with only an intercept, intercept and linear slope and with an intercept, linear slope and quadratic slope. We also explored the growth factors by constraining and varying their variances. Specifying an intercept, linear and quadratic slope for growth factors provided the best model fit. The variances of the growth factors were found to be negligible when allowed to vary; therefore, they were constrained to zero.

The most appropriate number of latent classes was selected by examining model fit statistics (Bayesian information criterion (BIC), Akaike information criterion (AIC), Lo-Mendel-Rubin (LMR), Likelihood Ratio Test, and entropy values). External generalisability factors—including the sample size of the smallest class and how classes aligned to previous literature—were also considered. To account for missing data, a full information maximum-likelihood approach was used; this approach estimates a likelihood value for each individual based on their existing data, which is then weighted and summed for all cases to determine model parameters.28

Using the model developed from objective 1, psychosocial work conditions, sex and all other covariates were regressed on participants’ trajectory class membership using a multinomial logistic regression to determine the association between psychosocial work factors and alcohol consumption trajectories (objective 2).

To answer our third objective, a cross-product term between each psychosocial work factor and sex were computed. The original multinomial model from objective 2 was then stratified by sex to determine how each psychosocial work factor affects trajectory membership in men and women.

For all analyses, a sampling weight was used to account for initial probability of selection and non-response to the original NPHS survey. To account for the clustered survey design of NPHS, variances for each point estimate were estimated using a bootstrap procedure involving 500 replications for each model. Trajectory building was completed in Mplus V.7 (Muthén & Muthén, Los Angeles, California, USA), while data preparation and multinomial logistic regressions were explored in SAS V.9.4 (SAS Institute, Cary, North Carolina, USA).


Participant characteristics

Table 2 highlights the demographic, employment and health characteristics of the study sample. The sample included 2594 women and 2864 men. The average age in 1994 was approximately 37 years (SD=11.25). The majority of the sample were Caucasian (90.14%), born in Canada (81.63%), did not have health restrictions (87.04%) and had full-time employment (85.47%).

Table 2

Participant characteristics in1994

Objective 1

The model with the best fit was a three-latent class model (figure 1). This model had adequate entropy (0.804) and a statistically significant LMR test (P<0.0001), suggesting that this model provided more information about daily drinking consumption patterns compared with the two-class model. The LMR test was not statistically significant for the four-class or five-class models (online supplementary table 1), suggesting that they did not provide any more information regarding the population’s daily drinking consumption patterns compared with the three-class model.

Supplementary file 1

Figure 1

Latent alcohol consumption trajectories.

The first and largest class (n=2917) was composed of individuals reporting 0 drinks per day over the 16-year period. The second class was composed of approximately 2040 participants who reported drinking 0.5–1 drinks per day. In this class, less than 5% met Centers for Disease Control and Prevention (CDC) criteria for binge drinking (four or more drinks per day for women and five or more for men).29 Roughly 40% of this class met the CDC criteria for ‘heavy drinking’ (8 or more drinks per week for women and 15 or more for men), with the majority of these individuals being female.30 The final class (n=502) reported drinking 2–3 drinks per day. In this class, roughly 40% of the participants met binge drinking criteria and roughly 96% met heavy drinking criteria. These three classes will be referred to as non-drinkers, light drinkers and moderate drinkers, respectively. Generally, the trajectories were stable over the 16 years.

Objective 2

When regressing participants’ trajectory class membership on psychosocial work conditions, sex and covariates (table 3), it was seen that men had a higher probability of being in the light and moderate drinking classes. The odds for men to be in the light drinker class compared with the non-drinking class was approximately three times greater (95% CI 2.52 to 3.53) than for women. The odds of men in the moderate drinker class was approximately 14 times greater (95% CI 9.85 to 20.89) compared with women.

Table 3

Multinomial logistic regression of latent class membership on psychosocial work factors, sex and covariates*

Psychosocial work dimensions of physical exertion and social support were associated with the moderate alcohol consumption trajectory. For each unit increase on the physical exertion scale, the odds of being in the moderate drinking class compared with the non-drinking class increased by 12% (95% CI 1.01 to 1.23), suggesting that individuals with more physically demanding jobs were more likely to be in the moderate drinking class compared with the non-drinking class. For each unit increase on the social support scale, the odds of being in the moderate drinking class compared with the non-drinking class increased by 8% (95% CI 1.01 to 1.15), suggesting that individuals in this class reported less social support at their work.

Objective 3

When exploring psychosocial work factor and sex cross-product interactions, significant interactions were seen between psychological demands and sex for membership in the moderate drinking class (OR=0.79, 95% CI 0.63 to 0.99) and physical exertion and sex for both the light drinking (OR=1.26, 95% CI 1.10 to 1.43) and moderate drinking classes (OR=1.47, 95% CI 1.06 to 2.05) compared with the non-drinking class.

Table 4 presents the estimates of the multinomial regression models, stratified by men and women. Higher workplace physical exertion was associated with higher odds of being in the light drinking among men (compared to the non-drinking group), with a protective effect among women. Differences were also observed between men and women in the effect of psychological demands and moderate drinking, with psychological demands associated with an increased risk of membership in this group among women (although not statistically significant), but a decreased risk of being in this group among men (OR=0.91, 95% CI 0.83 to 0.99) while compared to the non-drinking group.

Table 4

Multinomial regression model of trajectory membership on psychosocial work factors, stratified by sex


This study aimed to determine the number of latent alcohol consumption trajectories among the Canadian working population over a 16-year period and whether psychosocial working conditions were associated with drinking behaviour among men and women. Three distinct trajectories were found within our sample: non-drinkers, light drinkers consuming roughly 1 drink per day and moderate drinkers consuming 2–3 drinks per day. These latent trajectories generally align to known alcohol use patterns among high-income nations.2 31 Our findings indicate that the majority of the Canadian working population are in alignment with low-risk drinking guidelines,32 with those in the highest alcohol consumption trajectory potentially at risk of heavy drinking. Very little variation in alcohol use was seen over the study period, suggesting that the sample did not change their consumption behaviours over time; this is consistent with stable drinking behaviours in adulthood seen in other longitudinal cohorts.33 34

Aligning to previous literature, we observed that lower reported levels of work social support and higher levels of physical exertion were associated with higher alcohol use. For example, Colell and colleagues12 found that among men surveyed in the Household Survey on Alcohol and Drugs in Spain, a lack of social support at work was associated with heavy drinking (OR=1.62). Marchand and colleagues21 also saw that physical demands were associated with high-risk alcohol consumption using cross-sectional Canadian data.

Two main mechanisms of how psychosocial work factors specifically impact alcohol use have been previously proposed: the tension-reduction model35 and the biphasic self-medication model.13 The tension-reduction model suggests that alcohol use is a coping mechanism to reduce negative feelings that occurs from experiencing work stressors.14 35 This theory has been expanded by Frone13 in the biphasic self-medication model, which suggests that individuals consume alcohol to mitigate the impact of negative affect (feelings of emotional distress, anxiety or irritability) and work fatigue caused by work exposures.

Although previous literature has shown other psychosocial exposures included in our study (eg, skill discretion, job insecurity)9 19 to be associated with alcohol use, these factors may not have been statistically significant in our study given that daily alcohol use did not vary within and across time points, or perhaps because all psychosocial factors were included in the same model which reduced the variability between associations. To explore this further, we completed a post hoc exploration of models where only one work stressor (rather than all six) was included and regressed on drinking trajectory class; we did not find any discrepancies with our presented results.

For our final objective, higher levels of physical exertion at work were association with the heaviest alcohol use among men, but lower alcohol use among women; this finding is consistent with previous cross-sectional studies.21 There are multiple reasons why men with physically demanding work may be more likely to engage in moderate-to-heavy alcohol behaviours, while women in similar work do not. Women in these occupations may consume less alcohol since it may impact their job performance more greatly than men.21 Since men generally have more body weight, they may be less likely to suffer from the neurological and hangover impairments associated with drinking compared with women,3 13 36 allowing for greater daily alcohol consumption. Second, men and women may have different coping strategies to deal with stressors in the workplace.11 According to the biphasic self-medication model, men may externalise distress more than women.13 Further, men may not perceive the social consequences of greater alcohol consumption and might perceive alcohol consumption as part of their gender identity.8 13 36 37 Lastly, women with healthier lifestyles (ie, increased physical activity, reduced alcohol and smoking consumption, etc) may select more physically demanding employment compared with women with less healthier lifestyles, while this selection process may not be present among men.38

Our stratified models also suggested that men with higher workplace psychological demands may drink less, while women may drink more. This finding for men in the workforce has been previously seen,39 but our results may dispute previous studies finding no association between psychological demands and problematic alcohol use among women in the workforce.9 21 40–42 Many previous cohorts included more men then women; although our effect estimate was not significant in the female moderate drinking model, this may be because the moderate drinking class was primarily male.43 This perhaps suggests that our and previous studies have been underpowered to detect an association between psychological work conditions and alcohol use among women.

The finding that women in psychologically demanding workplaces may engage in greater drinking behaviours compared with men may be because women in management, leadership or male-dominated work roles may engage in more drinking to match the drinking behaviours of their male colleagues.44 45 Further, in many industries, women in high-level management or leadership positions are generally the minority compared with men, which has been shown to be associated with additional work stress.46 Alternatively, perhaps these women perceive their roles as more hectic or are required to manage conflicts with the individuals they work with, with alcohol use being a coping mechanism.13 Additionally, many women in the Canadian workforce also have societal roles outside of the workforce (eg, taking care of their families), which may also be psychologically demanding.47 48 The combination of stresses from their home lives and from their workplace may have led women in the sample to engage in moderate alcohol consumption, which is a relatively socially acceptable behaviour in Canada. However, it should be emphasised that given that the moderate drinking class was composed of primarily men, further work exploring the association between the psychosocial work environment and alcohol use should be undertaken in female-focused samples.

Strengths and limitations

First, given the need for more complex and rigorous longitudinal methods to study the association between psychosocial working conditions and alcohol use, our use of latent class modelling is a novel contribution to the literature. Second, the use of a representative national sample of employed Canadians over a 16-year period allowed us to examine these relationships over a variety of industries, occupations and demographic groups.

However, this study also has limitations. First, the assessments of both alcohol use and psychosocial work factors were self-reported, potentially introducing measurement bias and most likely underestimating the true effects. Second, the workforce reflected in our study may differ from the present study given that the cohort began in 1994 and was primarily Caucasian; therefore, this work should be replicated in more current and diverse samples. Further, given that our study examines the entire working population of Canada, the present drinking trajectories may be different for specific occupation groups. Therefore, future research should replicate these findings in different work contexts, as they may be occupation or industry specific.

Third, although using average daily alcohol use as our dependent variable allowed us to align results with definitions for binge drinking and risky drinking, it could not allow us to explore alcohol dependence disorders or differences in weekday versus weekend drinking patterns. This information may also be of interest to policy makers and workplaces; therefore, we recommend that further research be done using different alcohol metrics to replicate or further these results. Fourth, our results do not take into account other health behaviours, such as body mass index, smoking status and physical activity, as these factors may be mediators of the psychosocial working conditions and alcohol use association; however, we did explore the influence of these variables in sensitivity analyses, where our results did not differ from those presented. Fifth, this study focused on how the psychosocial work environment in 1994 influenced alcohol use trajectories up until 2016; the relationship between alcohol use and the working environment may be bidirectional, which should be explored in future literature. Lastly, our analyses did not explore how coping mechanisms or previous mental health may affect the association between psychosocial working conditions and alcohol use, which may overestimate our results.

Conclusions and future directions

Our study revealed that there are currently three drinking behaviour trajectories among the employed Canadian population. Findings suggest that low social support and high physical exertion at work may be important predictors of subsequent alcohol use. In addition, physical exertion and psychological demands within the workplace may influence alcohol use differently among men and women. Future research should explore the associations between psychosocial work factors and other health behaviours using latent class modelling, as well as explore more culturally diverse and female samples.

What is already known on this subject

Although the influence of psychosocial working conditions on alcohol use has been previously studied, results have been mixed and inconclusive. Many studies do not take into account how psychosocial work conditions may influence alcohol behaviours over time, use stark dichotomies to define alcohol consumption or focus on male-dominated samples. Modelling the association between psychosocial working conditions and alcohol consumption trajectories has not yet been studied. This examination would contribute to the current national and international prevention efforts aimed at reducing heavy alcohol consumption.

What this study adds

Using latent group-based trajectory modelling, three distinct and stable alcohol consumption patterns (non-drinkers, light drinkers and moderate drinkers) among the employed sample were seen. High workplace physical exertion was associated with membership in heavier alcohol drinking trajectories among men but associated with membership in the non-drinking trajectory among women. Conversely, lower levels of workplace psychological demands among men but high levels among women were associated with membership in heavier alcohol consumption trajectories.


This work was supported through a project grant from the Canadian Institutes for Health Research (CIHR) (grant number 310898). PS is supported through a Research Chair in Gender, Work & Health from CIHR. MG-O is supported through a postdoctoral fellowship through CIHR. Access to the data for this paper was enabled through Statistics Canada Research Data Centre at the University of Toronto.



  • Contributors KGD aided in conceiving the study and its design, analysed and interpreted the data, drafted the initial manuscript and approved the final manuscript as submitted. SI aided in conceiving the study and its design, the analysis and interpretation of the data and reviewing and revising the manuscript. MG-O aided in interpreting the data, reviewing and revising the manuscript. CAM conceived the study and its design and aided in interpreting the data, reviewing and revising the manuscript. PS conceived the study and its design, aided in analysing and interpreting the data, reviewing and revising the manuscript. All authors participated in approving the final version to be published and agreed to be accountable for all aspects of the work by ensuring that questions related to the accuracy or integrity of any part of the work were appropriately investigated and resolved.

  • Funding This study was funded by Canadian Institutes of Health Research (grant number:310898).

  • Competing interests None declared.

  • Ethics approval Statistics Canada and Health Sciences ethics committee at the University of Toronto.

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

  • Data sharing statement Currently, the data used for this study are not available for sharing. However, any questions about the results or data used in this study may be directed to the corresponding author.