Examining the relationship between the demand-control model and incident myocardial infarction and congestive heart failure in a representative sample of the employed women and men in Ontario, Canada, over a 15-year period

Abstract

Objectives

To examine the relationship between job strain and incident myocardial infarction and congestive heart failure in a representative population of men and women in Ontario, Canada, over a 15-year period.

Methods

A total of 14,508 respondents having provided responses to either the 2000/2001, 2002, or 2003 cycles of the Canadian Community Health Survey (CCHS) were aged 35 and older at the time and working. After removing respondents with pre-existing heart disease and missing data, our sample totaled 13,291 respondents. Responses were linked to administrative health care and hospitalization data to capture incident cases of myocardial infarction and congestive heart failure up to March 31, 2017. Job control and psychological demands were assessed using 5 items and 2 items respectively. A series of time-to-event regression models were run, adjusting sequentially for socio-demographic variables and health, other psychosocial work exposures, and health behaviours and body mass index.

Results

Over the study period, there were 199,583 person-years of follow-up (median follow-up: 15 years, 233 days). Higher incidence rates were observed for men (6.69 per 100 persons) than for women (2.77 per 100 persons). No clear relationship was observed for demand-control exposures and incidence of myocardial infarction and congestive heart failure in either men or women. After adjustment for socio-demographic factors, pre-existing health conditions, and other psychosocial exposures, the hazard ratio for high strain exposure (compared with low strain exposure) was 0.92 (0.46–1.84) for women and 0.75 (0.44–1.27) for men.

Conclusion

In this large prospective cohort in Canada, we observed no relationship between components of the demand-control model and incident myocardial infarction and congestive heart failure over a 15-year period.

Introduction

The psychosocial work environment has been proposed as a key mechanism linking economic and social structures to health and illness among the working population (Rugulies 2019; Smith et al. 2008). The psychosocial work environment refers to expectations of employees in regard to opportunities for learning, fairness in how they are treated, and positive interactions with others in the workplace (Marmot and Siegrist 2004). One of the most studied theoretical models of the psychosocial work environment is the demand-control model (Karasek and Theorell 1990). Under this model, the work environment most deleterious to health is that where high psychological demands are combined with low job control (job strain). Job strain has been linked to heart disease in multiple systematic reviews (Sara et al. 2018; Taouk et al. 2020; Theorell et al. 2016), and a meta-analysis of published and unpublished cohorts in Europe and the United Kingdom (Kivimaki et al. 2012).

Many previous studies on job strain and its components and cardiovascular outcomes are from Europe or the UK, with very few studies based on North American worker populations (Choi et al. 2015). A previous study, among respondents to the National Population Health Survey (NPHS) between 1994 and 2010, observed no relationship between decision authority and skill utilization (components of job control) and psychological demands with self-reported physician-diagnosed heart disease, captured every 2 years (Marchand et al. 2017). However, it is possible this null finding may be due to a healthy worker effect, given the outcome could only be captured among participants who continued to participate in the NPHS and who are likely in better health and have higher household income (Swain et al. 1999). In this paper, we address some of the attrition and self-reported outcome limitations of this previous work by examining the relationship between job strain and incident heart disease, captured using administrative healthcare data, in a representative population of workers in Ontario, Canada, over a 15-year period.

Understanding whether the relationship between psychosocial work exposures and chronic disease differs for men and women remains an under-researched area (Quinn and Smith 2018). Previous studies from Canada have observed that women exposed to low job control have an increased risk of diabetes compared with men (Smith et al. 2012), but that men exposed to low job control have an increased risk of hypertension compared with women (Smith et al. 2013). These male/female differences in the association between job strain and its components on diabetes and hypertension have also been observed in systematic reviews (Cosgrove, Sargeant, Caleyachetty et al. 2012; Gilbert-Ouimet, Trudel, Brisson et al. 2013; Sui et al. 2016). Given heart disease is a common subsequent outcome of both diabetes and hypertension, understanding whether the relationship between job strain and its components and heart disease differs for men and women is an important area of research. Although the most recent systematic reviews of the relationship between job strain and heart disease have not observed a difference in the relationship between men and women (Sara et al. 2018; Theorell et al. 2016), examining male/female differences in a population where previous differences between the components of job strain and diabetes and hypertension outcomes have been observed would add to this area. As such, a secondary objective of this paper is to examine differences in the relationship between job strain, the components of job strain, and incident heart disease for men as compared with women.

Methods

Study sample and design

This study used responses from Ontario respondents to the 2000/2001, 2002, and 2003 cycles of the Canadian Community Health Surveys (CCHS). Between 2000 and 2005, the design of the CCHS was to have large samples of approximately 130,000 respondents surveyed every 2 years, with smaller surveys (N approx. 35,000) examining specific areas in the years in between. The 2000/2001 and 2003 cycles of the CCHS are major cycles, where approximately 130,000 respondents were surveyed across Canada with a priority to give stable estimates from the survey within each of the approximately 135 health regions in Canada (i.e., a minimum sampling requirement is provided to each health region). The 2002 was a smaller sample of approximately 35,000 respondents where the sampling strategy was to provide stable estimates across 10 Canadian provinces (excluding Nunavut, Yukon, and the Northwest Territories).

For the purpose of this analysis, we focused on employed respondents, aged 35 and older, working 15 or more hours per week, who were in health regions that opted to include questions on the psychosocial work environment as part of their optional content. This sample totalled 14,508 respondents (N = 9186 from the 2000/2001 CCHS, 2996 from the 2002 CCHS, and 2326 from the 2003 CCHS). The differing number of respondents between the 2000/2001 and 2002 CCHS cycles reflects the number of respondents surveyed, while the difference between the 2000/2001 and 2003 CCHS cycles reflects differences in the number of health regions that opted to include the content on the psychosocial work environment (Smith et al. 2008). Of this sample, 3.3% were removed as they had pre-existing heart disease. An additional 455 respondents were missing responses for psychosocial work exposures, and 296 were missing responses for other exposures. After removing these respondents, the final analytical sample totaled 13,291 persons (95% of the sample free of heart disease at baseline). The study obtained ethics approval from the University of Toronto Health Research Ethics Board.

Primary outcome

Incident myocardial infarction and congestive heart failure over the follow-up period were derived using the Ontario Myocardial Infarction Database and the Ontario Congestive Heart Failure Database. Both these databases were developed using validated algorithms, with sensitivity and specificity estimates of approximately 0.85 or higher (Tu et al. 2003; Tu et al. 2001). These databases capture cases of myocardial infarction and congestive heart failure from 1992 up to March 31, 2017. For simplicity from this point forward, we refer to these conditions as heart disease, although we acknowledge the term heart disease constitutes more than just these two conditions. These datasets were linked to the CCHS responses using unique encoded identifiers and analyzed at ICES. ICES is an independent, non-profit research institute whose legal status under Ontario’s health information privacy law allows it to collect and analyze health care and demographic data, without consent, for health system evaluation and improvement.

Primary independent variables

In each cycle of the CCHS, an abbreviated measure of the job content questionnaire was used to assess aspects of the psychosocial work environment. This measure included five items on job control, two on psychological demands, two on co-worker support, and single items for supervisor support, physical demands, and job security. For the purpose of this analysis, we focused on job control and psychological demand items, as well as combinations of job control and demand. A previous study observed that the factor structure of the demand and control items in this abbreviated measure is similar for men and women (Bielecky et al. 2017). Cronbach’s alpha for the job control scale and for the psychological demands scale have previously been reported as 0.61 and 0.34 respectively (Wilkins and Beaudet 1998). The lower alpha estimates for each of these scales is likely due to a smaller number of items capturing a broad psychosocial dimension (Streiner 2003). Other psychosocial exposures were included in regression models. These were co-worker support (two questions), supervisor support (one question), job security (one question), and physical demands (one question). Details on the wording and reliability of these questions are available elsewhere (Wilkins and Beaudet 1998).

The four quadrants of the demand-control model were created using the median split approach for job control and psychological demand dimensions. These are: low strain (high job control and low demands); passive (low control and low demands); active (high control and high demands); and high strain (low control and high demands). In addition to the median split approach, we also subsequently defined demand-control categories using a job control and psychological demand quartiles, where we could create a fifth group of workers who are at the mid-point of the traditional demand-control categories (in the 2nd or 3rd quartile of both demands and control). This approach recognizes that many respondents are positioned around the median of the job control and psychological demands distributions and can address potential misclassification in demand-control categories.

Covariates

Covariates included age, sex, education, survey year, mode of interview (in person, by phone or a combination of both), work hours, weeks worked in the previous 12 months, marital status, presence of children less than 12 in the household, immigrant status, ethnicity (white versus other), presence of diabetes, hypertension, arthritis, back pain, other chronic conditions, and whether the person had activity restrictions that limited the type or amount of activity they could do at work. Health behaviours (smoking, alcohol consumption, and leisure-time physical activity) and body mass index were treated differently from other covariates, as it is unclear if these measures are on the causal pathway between work conditions and heart disease (mediators) or confounders. As such, they were included separately in regression models, as inclusion of these variables along with other covariates may result in over-adjustment (Schisterman et al. 2009).

Analyses

Initial analyses examined the distribution of all study variables and potential collinearity between covariates. Time-to-event regression models were used to examine the relationship between demand-control exposures and incident heart disease. Given death is a competing event, and because it is possible that respondents may leave the province of Ontario over the follow-up period, we examined four different approaches to classify time to event, to examine whether censoring assumptions impacted our results. A first approach right censored respondents at the development of heart disease, death from other causes, or the end of the follow-up period (March 31, 2017). A second approach, based on the work of Fine and Gray (1999), did not right censor at death, assuming those who had died are still at risk of developing heart disease. A third and fourth approach replicated these two approaches above, but also right censored respondents if they had not had any interaction with the health system in Ontario for a 5-year period (proxying for the respondent having left the province). Estimates did not differ to a great extent under different approaches. All results in this paper are where respondents are right censored at the development of heart disease or at the end of the follow-up period (the second of the approaches described above).

Three nested regression models were run. The first model adjusted for all health, general work, and socio-demographic covariates, but not other psychosocial work exposures or health behaviours and body mass index. The second model additionally adjusted for other psychosocial work variables (co-worker and supervisor support, job security, and physical demands), with the third additionally adjusting for health behaviours and body mass index. Estimates for the relationship between demand-control exposures and risk of heart disease for male and female samples were compared post hoc using stratified models and comparing point estimates and standard errors around each point estimate, to determine whether estimates from the male and female samples were statistically different from each other (Allison 1999).

Expansion weights are provided for each cycle of the CCHS which take into account the initial probability of selection and non-response to the survey. The sum of the expansion weights totals the estimated population of Canada aged 12 and over at the time of each survey. To accommodate differences in the number of respondents and thus the weighting between the 2000/2001, 2003, and the 2002 CCHSs (Thomas and Wannell 2009), we created a relative weight (by dividing the expansion weight by the average weight) for each CCHS cycle, so that the sum of the weights equaled the number of survey respondents in each cycle. Due to the complex sampling approach of the CCHS, variances around incidence estimates and hazard ratios have been performed using 500 replicate bootstrap weights, provided by Statistics Canada, which consider the survey design procedures for each cycle. All analyses were performed in SAS V 9.4 (SAS Institute, 2017).

Results

Over the study period, we had 199,583 person-years of follow-up (median follow-up: 15 years, 233 days). Full information on the distribution of study covariates is available in Tables 3 and 4 in the Appendix. Table 1 presents the incident rates of heart disease per 100 persons over the study period across quartiles of job control and psychological demand, and across demand-control groups, for the full sample and separately for men and women. Estimates for the demand-control groups including the group of workers centered around the medians for control and demands were not appreciably different from the four groups created with the median split approach and are presented separately in Table 5 in the Appendix.

Table 3.

Distribution of categorical study variables and incidence of heart disease. Employed Ontario respondents to the 2000/2001, 2002, and 2003 cycles of the CCHS working 15 or more hours per week and free of cardiovascular disease at time of survey (N = 13,291)

Variable	N	% of sample	Incident heart disease (%)
Survey year
2000–2001	8347	60.0	4.8
2002	2816	20.2	4.5
2003	2128	15.3	4.0
Mode of interview
In person	5920	42.5	5.0
Telephone	7373	53.0	4.1
In person and telephone	628	4.5	5.6
Sex
Female	6331	45.5	6.7
Male	6960	50.0	2.8
Education
Less than secondary education	1632	11.7	8.2
Secondary education completed	3877	27.8	4.8
Post-secondary completed below bachelor degree	4663	33.5	4.9
Bachelor degree and higher	3749	26.9	1.8
Marital status
Single (never married)	1829	13.1	1.3
Married	10,496	75.4	4.8
Widowed/separated/divorced	1597	11.5	5.4
Has children less than 12 in household
Yes	4374	31.4	2.8
No	8917	64.1	5.8
Living location
Urban	11,269	81.0	4.7
Rural	2022	14.5	5.5
Immigrant status
Less than 10 years	895	6.4	3.4
More than 10 years	3138	22.5	5.4
Non-immigrant	9258	66.5	4.5
Ethnicity
White	10,988	78.9	4.8
Non-white	2494	17.9	3.0
Missing	439	3.2	3.3
Hypertension
Yes	1567	11.3	10.3
No	11,724	84.2	4.1
Diabetes
Yes	456	3.3	12.6
No	12,835	92.2	4.6
Arthritis
Yes	1956	14.1	6.7
No	11,335	81.4	4.5
Back problems
Yes	2793	20.1	5.5
No	10,498	75.4	4.6
Other chronic condition
Yes	3662	26.3	5.3
No	9629	69.2	4.6
Activity restriction at work
Yes	2703	19.4	6.9
No	10,588	76.1	4.3
Alcohol consumption
Non-drinker	2119	15.2	5.9
Non-binge drinker	7117	51.1	3.8
Binge drink less than once per month	2575	18.5	4.5
Binge drink once a month or more	2110	15.2	5.1
Smoker
Yes	3554	25.5	7.4
No	9737	69.9	3.9
Leisure-time physical activity
Active	3160	22.7	2.7
Moderately active	3309	23.8	3.6
Inactive	7453	53.5	5.6
Body mass index
Underweight	135	1.0	(supp)
Normal weight	5148	37.0	2.0
Overweight	5438	39.1	4.6
Obese	3200	23.0	8.1

Table 4.

Average values for continuous study variables among populations with and without incident heart disease over the study period. Employed Ontario respondents to the 2000/2001, 2002, and 2003 cycles of the CCHS working 15 or more hours per week and free of cardiovascular disease at time of survey (N = 13,291)

Variable	With incident heart disease	Without incident heart disease
Age	50.8	45.5
Usual hours of work	42.9	42.2
Weeks of work in last 12 months	49.4	49.9

Table 1.

Incidence and corresponding 95% confidence intervals of cardiovascular disease by job control, psychological demand quartiles and demand control groups, for the full sample and separately for men and women. Employed Ontario respondents to the 2000/2001, 2002, and 2003 cycles of the CCHS working 15 or more hours per week and free of cardiovascular disease at time of survey(N = 13,291)

	All (N = 13,291)			Women (N = 6331)			Men (N = 6960)
	N1	Inc	95% CI	N	Inc	95% CI	N	Inc	95% CI
	13,291	4.82	4.27, 5.37	6331	2.77	2.28, 3.26	6960	6.69	5.75, 7.64
Job control quartiles
Highest	3939	4.70	3.37, 6.03	1597	2.55	1.54, 3.55	2342	6.17	4.07, 8.26
Q2	3636	4.75	3.79, 5.71	1664	2.23	1.29, 3.17	1972	6.88	5.29, 8.46
Q3	2683	5.51	4.19, 6.83	1403	3.02	1.84, 4.19	1280	8.25	5.85, 10.65
Lowest	3033	4.46	3.52, 5.39	1667	3.30	2.39, 4.21	1366	5.87	4.25, 7.48
Psychological demands quartiles
Lowest	2816	5.65	4.15, 7.15	1198	3.82	2.34, 5.30	1617	7.01	4.64, 9.38
Q2	3647	5.54	5.32, 6.76	1672	2.86	1.92, 3.81	1975	7.81	5.74, 9.87
Q3	4529	4.30	3.40, 5.20	2182	2.57	1.76, 3.38	2347	5.90	4.39, 7.42
Highest	2299	3.70	2.78, 4.62	1279	1.99	1.12, 2.85	1021	5.84	4.10, 7.58
Demand-control model
Low strain	3342	5.69	4.21, 7.17	1257	2.93	1.66, 4.19	2085	7.36	5.08, 9.63
Passive	3121	5.48	4.33, 6.63	1613	3.53	2.48, 4.58	1507	7.57	5.48, 9.66
Active	4233	3.96	3.06, 4.86	2005	2.05	1.30, 2.79	2229	5.68	4.11, 7.25
High strain	2595	4.32	3.26, 5.38	1456	2.78	1.72, 3.83	1139	6.29	4.32, 8.26

¹All estimates are weighted to take into account the probability of selection and non-response to each cycle of the CCHS

Inc incidence; CI confidence intervals

Table 5.

Incidence and corresponding 95% confidence intervals of cardiovascular disease by job control, psychological demand quartiles and demand-control groups including a fifth group of respondents centered around the median of the job control and psychological demand distributions (mid-population). Employed Ontario respondents to the 2000/2001, 2002, and 2003 cycles of the CCHS working 15 or more hours per week and free of cardiovascular disease at time of survey (N = 13,291)

	All			Women			Men
	N	Inc	95% CI	N	Inc	95% CI	N	Inc	95% CI
	13,291	4.82	4.27, 5.37	6331	2.77	2.28, 3.26	6960	6.69	5.75, 7.64
Demand-control (including mid-population)
Low strain	2204	5.83	3.81, 7.85	755	3.76	1.80, 5.72	1449	6.91	3.97, 9.85
Passive	2237	5.12	3.99, 6.25	1220	3.55	2.36, 4.75	1116	6.83	4.92, 8.73
Active	3024	3.91	2.79, 5.04	1432	2.04	1.09, 2.99	1592	5.60	3.63, 7.57
High strain	1630	3.84	2.67, 5.02	967	2.83	1.70, 3.97	663	5.32	3.06, 7.57
Mid-population	4097	5.17	4.15, 6.19	1957	2.39	1.56, 3.22	2140	7.71	5.95, 9.48

Inc incidence; CI confidence intervals

The incidence of heart disease was 4.82 per 100 persons in the full sample, with higher rates observed for men (6.69 per 100 persons) than for women (2.77 per 100 persons). No clear pattern of elevated risk of incident heart disease was present across quartiles of job control for the total sample, or for men and women. Across psychological demand quartiles, incidence of heart disease decreased as demands increased among the full sample, and for men and women. When demands and control were combined, the highest incidence of heart disease was observed for respondents in low strain and passive occupations and the lowest for active occupations.

Table 2 presents the regression estimates for demand-control groups with incident heart disease after adjustment for socio-demographic and health measures (model 1), additional adjustment for other psychosocial work measures (model 2), and additional potential overadjustment for health behaviours and body mass index (model 3). No statistical difference was observed in hazard ratios for passive, active, or high strain work environments (compared with low strain environments), for the full sample, and for men and women. Among both men and women, estimates for high strain work showed a protective, although not statistically significant, effect. No statistical differences were observed in estimates among women compared with among men (results available on request). Results for models examining quartiles of job control and psychological demands as separate dimensions are available in Table 6 in the Appendix.

Table 2.

Hazard ratios and corresponding 95% confidence intervals for dimension of the demand-control model and incident heart disease over an approximate 15-year follow-up. Employed Ontario respondents to the 2000/2001, 2002, and 2003 cycles of the CCHS working 15 or more hours per week and free of cardiovascular disease at time of survey (N = 13,291)

	Model 11		Model 22		Model 33
	HR	95% CI	HR	95% CI	HR	95% CI
All
Low strain	ref		ref		ref
Passive	0.99	0.68–1.42	0.97	0.68–1.39	0.98	0.67–1.41
Active	0.90	0.64–1.27	0.86	0.61–1.21	0.92	0.66–1.29
High strain	0.86	0.58–1.28	0.81	0.54–1.22	0.85	0.57–1.27
Women
Low strain	ref		ref		ref
Passive	1.05	0.60–1.85	1.08	0.61–1.91	1.00	0.58–1.72
Active	0.79	0.43–1.46	0.77	0.42–1.42	0.79	0.43–1.44
High strain	0.89	0.46–1.73	0.92	0.46–1.84	0.86	0.45–1.65
Men
Low strain	ref		ref		ref
Passive	0.99	0.63–1.55	0.95	0.60–1.49	0.98	0.62–1.55
Active	0.94	0.62–1.42	0.88	0.58–1.33	0.86	0.64–1.42
High strain	0.83	0.50–1.37	0.75	0.44–1.27	0.82	0.50–1.36

¹Model 1 adjusted for age, work hours, weeks worked, highest level of education, year of survey, children less than 12 years of age in household, rural or urban living location, marital status, immigration status, ethnicity, sex (for full model), presence of chronic conditions, interview method, and activity restrictions at work

²Model 2 additionally adjusted for co-worker support, supervisor support, job security, and physical demands

³Model 3 additionally adjusted for smoking, alcohol consumption, leisure-time physical activity, and body mass index

HR hazard ratios; CI confidence intervals

Table 6.

Hazard ratios and corresponding 95% confidence intervals for quartiles of job control and psychological demands and incident heart disease over an approximately 15-year follow-up. Employed Ontario respondents to the 2000/2001, 2002, and 2003 cycles of the CCHS working 15 or more hours per week and free of cardiovascular disease at time of survey (N = 13,291)

	Model 1		Model 2		Model 3
	HR	95% CI	HR	95% CI	HR	95% CI
All
Job control
Highest	ref		ref		ref
Q2	0.87	0.60–1.27	0.86	0.60–1.25	0.88	0.61–1.27
Q3	1.06	0.71–1.59	1.06	0.70–1.59	1.06	0.70–1.60
Lowest	0.77	0.50–1.18	0.74	0.47–1.15	0.75	0.49–1.14
Psychological demands
Lowest	ref		ref		ref
Q2	1.08	0.76–1.55	1.06	0.73–1.54	1.12	0.78–1.61
Q3	0.92	0.64–1.32	0.87	0.60–1.25	0.95	0.67–1.36
Highest	0.91	0.59–1.40	0.83	0.52–1.33	0.95	0.62–1.45
Women only
Job control
Highest	ref		ref		ref
Q2	0.71	0.37–1.38	0.72	0.38–1.38	0.68	0.35–1.31
Q3	0.94	0.52–1.69	0.95	0.53–1.73	0.88	0.48–1.62
Lowest	0.85	0.46–1.56	0.89	0.48–1.65	0.78	0.42–1.43
Psychological demands
Lowest	ref		ref		ref
Q2	0.77	0.44–1.33	0.78	0.44–1.37	0.70	0.39–1.25
Q3	0.73	0.40–1.34	0.73	0.39–1.37	0.70	0.38–1.26
Highest	0.62	0.30–1.28	0.60	0.29–1.24	0.60	0.29–1.24
Men only
Job control
Highest	ref		ref		ref
Q2	0.95	0.61–1.48	0.93	0.60–1.46	0.96	0.61–1.50
Q3	1.13	0.69–1.86	1.11	0.68–1.83	1.15	0.70–1.90
Lowest	0.73	0.42–1.27	0.67	0.38–1.20	0.72	0.42–1.24
Psychological demands
Lowest	ref		ref		ref
Q2	1.19	0.76–1.86	1.16	0.73–1.85	1.26	0.82–1.95
Q3	0.96	0.62–1.50	0.89	0.57–1.41	1.00	0.65–1.54
Highest	1.05	0.63–1.78	0.94	0.52–1.70	1.10	0.67–1.83

HR hazard ratios; CI confidence intervals

Discussion

Heart disease continues to be one of the leading causes of morbidity and mortality in developed economies (GBD 2016 DALYs and HALE Collaborators 2017). The objective of this study was to examine the relationship between job strain and incident myocardial infarction and congestive heart failure (referred to as heart disease) in a representative population of workers in Ontario, Canada, over a 15-year period, particularly focusing on potential differences between men and women. While we observed that the overall incidence of heart disease was higher among men than among women, no clear relationship was observed between categories of the demand-control and incidence of heart disease among either men or women. In addition, estimates for men and women were similar, despite previous studies having documented differences in the relationship between similar measures of the psychosocial work environment and diabetes and hypertension (two strong determinants of heart disease) in the same population previously (Smith et al. 2012; Smith et al. 2013).

The protective relationship between job strain and heart disease (compared with the referent group of low strain work) was not expected based on the bulk of the previous research in this area, although as noted in the Introduction, a previous Canadian study also observed no relationship between the components of job strain and self-reported heart disease measured on a 2-year basis (Marchand et al. 2017). However, another study from Quebec observed a twofold risk of recurrent myocardial events with chronic job strain (compared with no job strain), after a washout period of 2.2 years, among 972 respondents who returned to work after a first myocardial infarction (Aboa-Eboule et al. 2007).

Focusing on other studies in North America, there have been two population-based cohorts in the United States where the relationship between the demand-control model and incident heart disease has been examined. A 2004 study, using the Framingham Offspring Cohort, observed that active, low strain, and passive work were protective (although not statistically significantly) of heart disease among men (compared with high strain work), over a 10-year follow-up period. However, in this same study, a statistically significant elevated risk of heart disease was observed among women in active occupations (compared with high strain occupations), with an elevated but not statistically significant risk also observed among women in low strain occupations (Eaker et al. 2004). An earlier study, focusing only on 3575 male respondents to the National Health and Nutrition Survey I, observed no relationship between job strain and heart disease over a 12–16-year period (Steenland et al. 1997). Comparisons with other studies, in general, would be simpler if a more objective cut-point for job strain was available, rather than the continued reliance upon sample-based approaches (Smith and LaMontagne 2015).

Potential explanations for our null findings are that exposure to job strain either is not deleterious in relation to heart disease among Canadian workers or is not as relevant as other psychosocial work exposures, such as effort-reward imbalance, quality of leadership, vertical trust, or organizational justice, in modern labour markets. Recent work in Canada has observed that there have been limited changes in job control and psychological demands in Canada over the period 2002 to 2012 (Fan and Smith 2018). The inconsistency of this observation with anecdotal reports of how the nature of work is changing in Canada may suggest that the most important dimensions of work are not captured by the abbreviated job content questionnaire used in our study. However, as noted, the job control component of this abbreviated measure has been associated with an increased risk of diabetes (Smith et al. 2012) and hypertension (Smith et al. 2013). These findings suggest that the abbreviated measure of psychosocial work exposures used in this study does have important associations with other health conditions.

It is also potentially possible that there is something unique about the publicly funded health care system in Ontario that impacts the relationships between cardiovascular risk factors and outcomes. However, we did observe expected relationships between age, education level, sex, pre-existing diabetes, pre-existing hypertension, smoking status, and body mass index and risk of heart disease in our final regression models (results not shown, but available on request).

Strengths and limitations of the study

The novel linkage between a population-level survey with a relatively low non-response rate, capturing information on a variety of aspects of the psychosocial work environment, with population-level administrative data capturing health care services provided in a public health care system, allowed a unique examination of the relationship between dimensions of the psychosocial work environment and heart disease over a relatively long 15-year period. Our population-based sample also resulted in good variation across psychosocial work measures, with members of our sample observed across all possible job control and psychological demand values. The capture of cardiovascular events was done with validated algorithms, and the available linkage allowed the removal of respondents with previous cardiovascular events, even if these were not reported as part of the baseline survey. Although the CCHSs capture a variety of information about an individual’s socio-demographic characteristics, their health status and restrictions in activity and health behaviours, and aspects of their work environment, all this information is measured cross-sectionally. As a result, it is not possible to be sure of the temporal relationships between various measures included in our analyses (e.g., work environment and health behaviours, or work environment and other health measures). In addition, the assessment of the psychosocial work environment was based on an abbreviated measure, administered at one time point. Having additional measures of the work environment over time would strengthen the current study, as we would be able to incorporate more information into the length of time respondents had particular psychosocial work exposures. Finally, we did not have access to clinical variables such as blood pressure readings, cholesterol levels, or medications.

Conclusion

In the current study, we observed no clear relationship with different demand-control psychosocial exposures and incident heart disease measured over a 15-year follow-up period. Future work should explore the continued relevance of the demand-control model in contemporary labour markets; and if the demand-control model is still relevant, the particular aspects of the Ontario labour market context that might lead to the null relationships observed in this paper.

Author contributions

PS led the design of the study and hypothesis generation with contributions from all study authors. PS wrote the first draft of the manuscript, and all authors contributed to the interpretation of the data and critical review of the paper and approved the final version.

Funding information

This work was supported by the Canadian Institutes of Health Research (CIHR). RG was supported as a Clinician Scientist in the Department of Family and Community Medicine at the University of Toronto and St. Michael’s Hospital. MGO held a CIHR research grant for her postdoctoral fellowship. PS was supported by a Research Chair in Gender, Work and Health from CIHR. This study was supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care (MOHLTC). Parts of this material are based on data and information compiled and provided by the MOHLTC and the Canadian Institute for Health Information (CIHI).

Compliance with ethical standards

The study obtained ethics approval from the University of Toronto Health Research Ethics Board.

Conflict of interest

The authors declare that they have no conflict of interest.

Disclaimer

The analyses, conclusions, opinions, and statements expressed herein are solely those of the authors and do not reflect those of the funding or data sources; no endorsement is intended or should be inferred.