Aims: To examine whether job strain (ie, excessive demands combined with low control) is related to smoking cessation.
Methods: Prospective cohort study of 4928 Finnish employees who were baseline smokers. In addition to individual scores, coworker-assessed work unit level scores were calculated. A multilevel logistic regression analysis, with work units at the second level, was performed.
Results: At follow-up, 21% of baseline smokers had quit smoking. After adjustment for sex, age, employer and marital status, elevated odds ratios (ORs) for smoking cessation were found for the lowest vs the highest quartile of work unit level job strain (OR 1.43, 95% CI 1.17 to 1.75) and for the highest vs the lowest quartile of work unit level job control (OR 1.61, 95% CI 1.31 to 1.96). After additional adjustment for health behaviours and trait anxiety, similar results were observed. Further adjustment for socioeconomic position slightly attenuated these associations, but an additional adjustment for individual strain/control had little effect on the results. The association between job strain and smoking cessation was slightly stronger in light than in moderate/heavy smokers. The results for individual job strain and job control were in the same direction as the work unit models, although these relationships became insignificant after adjustment for socioeconomic position. Job demands were not associated with smoking cessation.
Conclusions: Smoking cessation may be less likely in workplaces with high strain and low control. Policies and programs addressing employee job strain and control might also contribute to the effectiveness of smoking cessation interventions.
Statistics from Altmetric.com
Smoking cessation is not a single event but a complex process that can be influenced by a wide range of factors, including work characteristics.1 The Karasek demand/control model of job strain postulates that psychological strain at work results from the interaction of job demands and job control, with the combination of low control and high demands producing job strain.2 3 Low control and high demands are also used as independent indicators of occupational stress. Earlier evidence suggests that high job strain could be predictive of morbidity and mortality, and it has also been associated with health-risk behaviours such as smoking status and smoking intensity.4–7
It is hypothesised that job strain conditions are predictive of smoking cessation. This is because stressors in the work environment can make it harder to quit, contribute to difficulties with cessation and possibly induce relapses.8 Smoking may be used either deliberately or inadvertently as a coping mechanism to deal with stress;9 almost all smokers attribute their smoking at least partly to its alleged calming and relaxing properties.10 Job strain could therefore strengthen a smoking pattern that is characterised by relief from a state of arousal that may have been produced or maintained by job strain.1 By contrast, resources in the work environment, such as high job control, might facilitate smoking cessation.11
Existing prospective evidence of the relationships between work characteristics and smoking cessation is scarce and mixed. A recent systematic review identified only four prospective studies on job strain and smoking cessation; none of them found a positive association.1 By contrast, there was some indication that high job demands and high resources (such as high job control) may be associated with a higher likelihood of cessation. Most of the earlier studies analysed small samples, have used inconsistent measures of job demands and control and did not differentiate between light and heavy smokers.
A further limitation in previous studies is the reliance on individual-level assessments of job strain and its components. However, differences in individual level job strain scores between respondents may not only reflect variation in the exposure to job strain between the respondents, but also the differences in response styles (eg, a personal disposition to answer negatively) between the respondents. The latter may cause bias in the estimation of the association between job strain and smoking cessation. A complementary approach would be to examine job strain using aggregated scores—for example by assigning a work unit mean score of job strain to each member of the work unit. This approach potentially safeguards against bias from randomly mismeasured individual level job strain.12
To extend the understanding of the relationship between job strain and smoking cessation, we examined this relationship in a heterogeneous cohort of Finnish public sector employees who were smokers at baseline. In addition to individual level scores, we used work unit level coworker-assessed scores to model the effect of job strain and its components in all baseline smokers and separately in light and moderate/heavy smokers. To the best of our knowledge, this is the largest prospective study on job strain and smoking cessation to date.
Sample and design
Data were obtained from the Finnish Public Sector Study, which is an ongoing prospective study examining the relation of behavioural and psychosocial factors with health. The study is focused on the entire personnel of 10 municipalities and 21 hospitals.13 Baseline survey data were collected between 2000 and 2002. The response rate was 68%. In 2004–2005, a follow-up questionnaire was sent to all identifiable respondents of the baseline survey who were still alive. A total of 35 914 responses were received (response rate 77%). Mean (SD) follow-up was 3.59 (0.81) years.
For this study, baseline non-smokers were excluded (n = 29 142). Employees with missing data on the size of their work unit (n = 5) or who worked in work units with less than three respondents (n = 148), were also excluded as work unit level job strain is only meaningful in groups. Moreover, participants categorised as baseline smokers but who reported smoking on average 0 cigarettes a day (n = 21) and participants with missing data for any study variables (n = 1670) were excluded. The final sample of this study included 4928 baseline smokers who also responded at follow-up.
Ethical approval for the study was obtained from the Ethics Committee of the Finnish Institute of Occupational Health.
Assessment of job strain was based on a modified job content questionnaire comprised of the job demand scale (Cronbach α = 0.76) and job control scale (Cronbach α = 0.82).3 14 Three questions addressed the psychological demands of the job, that is, having high workload and working at a high pace and not having enough time to complete work tasks. Job control was assessed with nine questions about the worker’s ability to use and develop skills and exert decision authority. The responses were given on a Likert scale of 1 = “very little” to 5 = “very much”. The mean scores for each of the two constructs were computed. To construct a continuous job strain measure, the means of job demand scores were subtracted from the means of job control scores.4 15–17
The dataset included individuals (employees) nested within work units in municipalities and hospitals. In addition to individual scores, we calculated aggregate-level job control, job demands and job strain scores according to the work units. We determined 1966 functional work units typically at a single location (eg, a school or a hospital ward) based on employers’ registers on administrative units used to allocate organisational resources, paying salaries etc. From the organisational hierarchies with multiple levels, the units at the lowest organisational level were selected, but only over two-person units were included. On this basis, the mean work unit size was 29 employees (range 3 to 319 employees). In general, work units were relatively small: 80% of the participants worked in work units with less than 35 employees. Only 5% of participants worked in work units with more than 100 employees. Aggregated job control, job demands and job strain of the work unit (second level) were calculated for each individual as the mean of his/her (first level) coworkers’ responses from the same work unit (self-estimation excluded).
Individual and work unit level scores were divided into quartiles for the analysis.
Smoking was measured at baseline and at follow-up using the following questions: “Do you smoke or have you previously smoked regularly, that is, daily or nearly daily?” and “If you have smoked, do you still smoke regularly?” and “How many cigarettes you smoke (or smoked) a day on average?” Smoking intensity at baseline was dichotomised (1 to 9 cigarettes and > = 10 cigarettes per day). Self-reported smoking status has been validated in several studies.18 19 As all employees included in this study were regular smokers at baseline, smoking cessation was detected if the participant reported that he/she was not a regular smoker at follow-up.
Nine covariates were included in the analysis. As smoking cessation is associated with various sociodemographic characteristics,20 the following characteristics were included: sex, age, marital status, employer (municipality vs hospital) and socioeconomic position (SEP) based on occupational-title classification of Statistics Finland; ie, higher-grade non-manual workers (eg, doctors, teachers), lower-grade non-manual workers (eg, technicians, registered nurses) and manual workers (eg, cleaners, maintenance workers). Sex, age, employer (municipality or hospital) and SEPs were obtained from the employers’ records, while marital status (married or cohabiting vs single, divorced or widowed) was obtained from survey responses.
In addition, we measured self-reported heavy drinking, physical activity and body mass index (BMI). BMI was calculated as self-reported weight (in kilograms) divided by self-reported height (metres) squared. Self-reported habitual frequency and amount of beer, wine and spirits intake was transformed into grams of absolute alcohol per week. One unit of pure alcohol (12 g) is equal to a 12-cl glass of wine, a single 4-cl measure of spirits, or a 33-cl bottle of beer. A dichotomous variable was created to represent heavy drinking, with a cut-off point corresponding to a weekly consumption of 210 g or more absolute alcohol and with all other respondents in the reference category.23
The participants reported the average amount of time spent per week on leisure and on the journey to and from work in physical activity corresponding to the activity intensity of walking, vigorous walking, jogging and running. The time spent at each activity was multiplied by its typical energy expenditure, expressed in metabolic equivalent tasks (METs).24 An activity MET index was expressed as the summary score of MET h/week.
According to the prerequisites of multilevel analysis, our dataset included participants nested within work units. We used the intraclass correlation coefficient (ICC) to study the differences of variance in job strain between work units. Technically, the multilevel ICC is a variation partition coefficient that indicates the proportion of the total variance of job strain that occurs at the work unit level.25 In our sample the ICC was 17%, indicating significant variance of individual job strain between work units.
We applied multilevel logistic regression analysis to examine the associations of individual and work unit level job strain, job control and job demands at baseline with smoking cessation at follow-up. The results are presented as adjusted odds ratios (ORs) and 95% CIs. The hypothetically least favourable conditions (the highest quartiles of job strain and job demands and the lowest quartile of job control, respectively) were selected as reference categories. Adjustments were made in steps in order to distinguish the different types of confounders. First: sex, age, marital status and employer; second: heavy drinking, BMI, physical activity and trait anxiety; and third: SEP. When modelling the associations of work unit strain, job control and job demands with smoking cessation, the corresponding individual level job strain variable was additionally adjusted for in the fourth model. In addition, the analyses were performed separately in light (<10 cigarettes per day) and moderate/heavy (> = 10 cigarettes per day) baseline smokers.
All statistical analyses were performed with SAS V. 9.1.3 statistical software (SAS Institute, Cary, North Carolina, USA) applying the Glimmix procedure. This procedure fits statistical models to hierarchical data with correlations or non-constant variability. It allows for simultaneous examination of the effects of the individual and group level variables on individual level outcome, while accounting for the non-independence of observations within groups.
The characteristics of 4928 baseline smokers are shown in table 1. Of the participants, 69% were married or cohabiting and more than half were in lower-grade non-manual positions. Approximately three-quarters of the participants were contracted to municipalities and the rest to hospitals.
The final cohort did not substantially differ from all baseline smokers (n = 5756) in terms of mean age (43.6 years in the sample vs 43.8 years in baseline smoker population), the proportion of women (77% vs 77%), SEP (28% manual vs 29% manual) and the mean of individual job strain at baseline (−0.41 (1.14) vs −0.41 (1.14)).
A total of 1031 baseline smokers (21%) quit smoking between baseline and follow-up. The quitters were on average younger, had lower BMI and lower level of trait anxiety. In addition, they were more often from higher SEP groups, worked for hospitals and were not heavy drinkers (table 1). Women, those living without a partner, hospital employees, manual workers and non-heavy drinkers experienced high job strain more often than their counterparts (p in all cases <0.001 except in marital status where p = 0.037 and heavy drinking where p = 0.031). Moreover, employees in the high strain group were on average older than the employees in other job strain categories (p = 0.013) and scored higher on trait anxiety (p<0.001) (data not shown).
Individual level job strain
Table 2 presents the results from the multilevel logistic regression models for individual level job strain variables and smoking cessation. After adjustment for sex, age, marital status and employer, employees in the lowest quartile of job strain had approximately 1.4-fold higher odds for smoking cessation compared with smokers in the highest quartile of job strain. Similarly elevated ORs of smoking cessation were obtained for the participants in the highest quartile of job control. These associations remained after health behaviours and trait anxiety were added into the models, whereas further adjustment for SEP attenuated the associations and they became insignificant. Job demands were not associated with smoking cessation in any of the models.
Coworker-assessed work unit job strain
Table 3 shows the results from multilevel logistic regression analyses on the associations of coworker-assessed (self-assessment excluded) work unit level job strain, job control and job demands with smoking cessation among baseline smokers. The results were largely similar to those obtained with individual level scores, although the associations with job strain and job control remained significant also after adjustment for SEP. After adjustment for sex, age, marital status, employer, heavy drinking, BMI, physical activity, trait anxiety, SEP and individual job strain/job control, low vs high work unit job strain was associated with 1.3 times higher odds and high vs low work unit job control was associated with 1.4 times higher odds of smoking cessation. Similarly as in individual level models, job demands were not associated with smoking cessation.
To examine whether large work units biased the results of coworker-assessed job strain, we re-ran the main analyses excluding those participants (n = 268) who worked in work units with 100 or more employees. The results remained after this exclusion. In the final model, ORs for low work unit strain and high work unit job control were 1.27 (95% CI 1.02 to 1.59) and 1.39 (95% CI 1.10 to 1.76), respectively. Low work unit job demands had no effect on smoking cessation, OR 1.09 (95% CI 0.87 to 1.36) (data not shown).
Analyses by baseline smoking intensity showed that the association between the lowest quartile of work unit job strain and smoking cessation was significant in light (<10 cigarettes per day) but not in moderate/heavy smokers (> = 10 cigarettes per day) (tables 4 and 5). However, in moderate/heavy smokers, elevated ORs for smoking cessation were observed in two intermediate categories of work unit job control.
There are three main results that arise from this prospective cohort study in Finnish public sector employees. First, low job strain and high job control were associated with a higher likelihood of smoking cessation at the aggregated work unit level, whereas at the individual level similar associations were observed only before but not after adjustment for SEP. Second, at the work unit level, the association of job strain with smoking cessation was slightly stronger in those baseline smokers who smoked less than 10 cigarettes per day. Third, job demands were not associated with subsequent smoking cessation. Age, sex, marital status, employer, heavy drinking, BMI, physical activity and trait anxiety could be confounders of the associations between the job strain variables and smoking cessation. Adjusting for these variables, however, only marginally affected the estimates.
With a sample size of almost 5000 baseline smokers, this is apparently the largest study of job strain and smoking cessation to date. Our results are in line with a 15-year follow-up study which found high control over work to be associated with an increased likelihood of smoking cessation.26 They are also in accordance with a finding that smoking was more likely be maintained after major organisational downsizing, a stressful situation at work associated with decrease in job control and increase in job demands, than after no downsizing.27 Moreover, our results confirm the results of Hibbard26 and Landsbergis et al,28 who did not find an association between job demands and smoking cessation. However, some other studies have found an association between different measures of demands and smoking cessation.29 30 We found an association between low job strain and smoking cessation whereas earlier studies have not reported such association.1 Potential reasons for these inconsistencies include (a) small sample sizes in many earlier studies possibly contributing to lack of statistical power and decreasing the likelihood of detecting significant associations; (b) differences in measurement of control and demands across studies; and (c) that the previous studies were based solely on individual-level assessments of job characteristics. It has also been hypothesised that certain kinds of demands might support smoking cessation under the right circumstances (eg, high control), while other kinds of demands can have an opposite effect.1 Furthermore, in general in earlier research a higher consistency of results has been obtained concerning the relation of low control with ill health outcomes such as cardiovascular disease, whereas the results have been more mixed with regard to the effect of demands.31
We found coworker-assessed work unit job strain to predict smoking cessation even after taking into account the effect of the baseline smokers’ own reports of job strain. There are at least three potential explanations for this finding. First, elevated levels of job strain in a work unit might decrease the amount of support given to those quitting smoking. Second, it is possible that in a work unit where there are a lot of people who experience job strain the employees are less likely to exercise control over adverse health behaviours such as smoking but rather maintain smoking approving culture. Third, work unit job strain could be a marker of other factors influencing propensity to stop smoking.
Our study is unique in showing a slightly stronger effect of work unit job strain on smoking cessation among light smokers than moderate or heavy smokers. We also found that younger employees were more likely to quit than older smokers. Both these results might relate to nicotine dependence, which is likely to be lower in younger and light smokers making them more responsive to positive environmental influences, such as low job strain. In contrast, high job strain could contribute especially to a smoking pattern that is characterised by relief from an arousal that may have been produced or maintained by job strain1 and thereby prevent quitting. However, the mechanisms by which stress functions to maintain smoking behaviour are complex and not well understood.10 There are also individual differences in the pattern of behavioural responses to adverse psychosocial factors, such as high job strain.32
Strengths and limitations
Our study was based on large prospective data on employees who were smokers at baseline. Response rates for baseline and follow-up surveys were satisfactory. The respondents represented the target population well in terms of mean age and the distribution by socioeconomic position. This reduced the potential for selection bias. We adjusted for multiple covariates, thereby limiting confounding bias. In addition, by using prospective data and complementing individual assessment of job strain with an ecological approach, the problems related to reporting and common-method variance bias were reduced.
However, our study is subject to several limitations. First, although the response rates were acceptable and there was little evidence of systematic non-response, non-response may still cause some bias. Moreover, we cannot totally rule out the possibility that continuing smokers have been selected into workplaces characterised by high job strain or low control. However, such a selective retention seems highly unlikely.
Second, self-reported measures of smoking were used without complementary data on more direct measures, such as serum cotinine levels. However, earlier studies have consistently shown that self-reported tobacco smoking is a valid and reliable way to measure smoking habits in a population.33 The proportion of smokers who deny or minimise smoking in survey studies may be negligible because they do not significantly change the results according to smoking status and self-reports of smoking has been found to be correlated with biological measures.19
Third, the Finnish versions of the job control and job demands measures were derived from the Job Content Questionnaire,3 but the demand scale in our survey was restricted to three items and, thus, was not identical with the original measure. Moreover, it did not include the measure of social support. It is possible that this has reduced the validity of the job strain assessment.
What this paper adds
Low coworker-assessed job strain and high coworker-assessed job control were associated with an increased likelihood of smoking cessation in a large sample of Finnish public sector employees who were baseline smokers
Job demands were not associated with smoking cessation
Policies and programs addressing employee job strain and control at work might support the effectiveness of smoking cessation interventions
Fourth, to complement analysis based on individual level scores, we used work unit aggregated scores of job strain as a means of controlling for bias arising from individual differences in response styles when modelling the effect of job strain variables on smoking cessation. This approach has potential weaknesses as the aggregate variables may be less sensitive as a measure of individual’s exposure to job strain than the individual level job strain scores are. However, having consistent findings from both of these approaches (each of which has distinct, but differing potential limitations) strengthens the evidence of the association between job strain and smoking cessation.
Fifth, although we took several confounding factors into account, there are other potential factors that we were not able to consider in the analyses. For example, smoking habits of a partner or friends or stressors at home might influence the associations between job strain variables and smoking cessation. Other unmeasured factors possibly related to smoking cessation, such as readiness to change,20 self-efficacy,34 and level of nicotine dependence35 could also confound the results if they are also related to job strain. In addition, it is possible that in some workplaces smoking had been totally banned, whereas in some other workplaces smoking may have been allowed in designated areas and this could have affected the results.
Sixth, our study did not contain data on whether baseline smokers that became non-smokers had any assistance or treatment to facilitate the cessation process. Some employers may have offered smoking cessation benefit for their employees. Providing smoking cessation benefit has been found to result in a greater number of successful cessations.36 However, such benefits would confound the findings only in case they systematically differ between workplaces with high vs low job strain and this seems unlikely.
Seventh, smoking cessation is a dynamic process that begins with a decision to stop smoking and ends with abstinence maintained over a long period of time.37 Attempting to quit smoking has two major components: initiating the attempt and maintaining cessation once quit.35 In the present study we do not know whether the reported smoking cessation resulted in a longer-term abstinence than the period of time covered by the study and how many times the smoking status of an individual changed between the two surveys.
Finally, the respondents were smokers from the Public Sector Study which is representative of Finnish public sector employees in terms of sex and age (77% female, mean age 44.6), but the female predominance does not correspond to the sex distribution of the Finnish general working population (48% female, mean age 45.5).38 Furthermore, with the cohort including a large proportion of hospital employees, health professional were over-represented. Therefore, the extent to which our results can be generalised to other populations is not known and the findings should be interpreted with caution until they are validated in studies using other samples.
Despite widespread public awareness of health risks of tobacco smoking, it remains the leading preventable cause of disease and death in Western societies39 and new evidence to assist the development of new approaches to reduce tobacco use would be very important. This prospective study found that certain aspects of psychosocial work environment, that is, low job strain and high job control, are related to increased likelihood of smoking cessation. The results suggest that “job control” rather than “job demand” factors may be more salient in relation to smoking cessation. As there is a lack of large-scale prospective employee cohort studies, our study contributes importantly to the literature on job strain and smoking cessation and highlights the importance of the work environment on health behaviours and thereby health and well-being of employees.
If causal, the observed associations can have important implications for workplace health promotion policies and interventions. Workplaces have increasingly become a channel for promoting smoking cessation. Large groups of smokers can be reached and non-smokers can be protected against second-hand smoke.40 The results of this study give tentative support for workplace policies and interventions to decrease job strain and increase job control in work units as a means of promoting smoking cessation. Integrated job stress and smoking intervention strategies have shown greater smoking cessation rates than traditional health promotion only interventions.41 42 Therefore studies of smoking interventions combined with interventions targeted to reduce job strain could be beneficial.
Competing interests: None declared.
Funding: The work presented in this paper was supported by grants from the Academy of Finland (projects 105195, 110451, 117604, 124271 and 124322), the Finnish Work Environment Fund (project 103432) and the participating towns and hospitals.
Ethics approval: Ethical approval for the study was obtained from the Ethics Committee of the Finnish Institute of Occupational Health.
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.