Employment quality and suicide, drug poisoning, and alcohol-attributable mortality

Abstract

Suicide, drug poisoning, and alcohol-attributable mortality (SDAM)—often labeled “deaths of despair”—are increasing among working-aged individuals in many high-income countries. We examined the association between employment quality and SDAM in Canada. Census records from the 2006 Canadian Census Health and Environment Cohort (n = 2 805 550) were linked to mortality data from 2006 to 2019. Latent class analysis identified 5 employment quality types: standard (secure and rewarding), portfolio (rewarding but demanding), marginal (limited hours and earnings), intermittent (sporadic and unstable), and precarious (insecure and unrewarding). Poisson regression models estimated sex/gender-stratified associations between employment quality type and SDAM separately. We observed a consistent mortality gradient across employment quality groups, with lower-quality employment—and precarious employment in particular—associated with increased rates of SDAM relative to higher-quality (ie, standard) employment. For example, precarious employment was associated with a more than 3-fold rate of drug poisoning deaths among women (rate ratio [RR] = 3.58; 95% CI, 3.21-4.00) and a more than 2-fold rate of alcohol-attributable death among men (RR = 2.22; 95% CI, 2.07-2.38). Employment quality is an important determinant of SDAM, with varying associations by sex/gender. Improvements in employment conditions may help reduce the burden of premature deaths attributable to suicide and substance use.

Introduction

Life expectancy is stalling in Canada and other high-income countries.^1-4 These trends predate the COVID-19 pandemic and are partly attributable to increasing suicide, drug poisoning, and alcohol-attributable deaths (SDAM)—often characterized as “deaths of despair.”^5-7 Rates of SDAM have increased against a backdrop of widening income inequality, declining labor market prospects, and a shrinking social safety net.^8-10 The concurrent nature of these trends has led researchers to posit a link between socioeconomic insecurity and SDAM.^11-14 Consistent with that view, evidence suggests that, within the working-age population, rates of SDAM are increasing most rapidly among socioeconomically disadvantaged groups.^8,15

Employment is a prominent theme in research on the socioeconomic determinants of SDAM.^8,12,16-19 An erosion of the “standard employment relationship,” characterized by full-time, permanent, year-round employment, and a resulting increase in precarious jobs (eg, temporary, insecure, low wage) are hypothesized to have played a crucial role in precipitating observed increases in SDAM.^20,21 Previous studies have documented an association between lower rates of labor force participation and higher rates of suicide and substance use–related harms.^13,22-24 Other studies have identified unemployment as an important risk factor for premature death, including SDAM.^15,25-28 At present, however, there is a limited understanding of how the quality of employment (as opposed to lack of employment) is related to SDAM.

Employment quality refers to aspects of an employment arrangement, such as hours, stability, and earnings, that influence health and well-being.²⁹ It is a multidimensional concept capturing a continuum of employment conditions, ranging from the “standard” ideal type of full-time, permanent, year-round employment at 1 end of the spectrum, to progressively more precarious forms of work at the other.^30-32 Whereas high-quality employment can be beneficial to health, extensive literature links low-quality (eg, precarious) employment to a range of adverse health outcomes, including higher rates of depression, suicidal ideation, substance use, psychological distress, and other mental health problems, that act as precursors to SDAM.^33-42 These findings highlight the potential salience of employment quality as an underlying driver of suicide and substance use–related death.

Despite a plausible connection between these phenomena, little is known about the impact of employment quality on SDAM. Furthermore, there is evidence that women and men experience disparate health and labor market outcomes and that the relationship between employment quality and health may vary by sex/gender.^37,43,44 Therefore, following a cohort of more than 2.8 million workers between 2006 and 2019, we estimate here the association between employment quality and rates of SDAM by sex/gender in Canada. In so doing, we highlight the role of employment quality as an upstream and potentially modifiable risk factor for SDAM.

Methods

Data and study population

We conducted a population-based cohort study using the 2006 Canadian Census Health and Environment Cohort (CanCHEC), a nationally representative data set comprising 5.9 million Canadians who completed the mandatory long-form Census in 2006.^45,46 The long-form Census is administered every 5 years to a random sample of 1 in 5 households and is designed to provide a portrait of the demographic, social, and economic characteristics of the household population in Canada. The CanCHEC cohort integrates long-form Census records with health administrative data, including mortality records from the Canadian Vital Statistics–Death Database (CVSD).⁴⁶ The CVSD is an administrative database of all deaths in Canada.⁴⁷ Record linkage was completed by Statistics Canada.⁴⁶ At present, the linkage includes all deaths that occurred in Canada between Census Day on May 16, 2006, and December 31, 2019. The response rate to the 2006 long-form Census was 93.8%.⁴⁶ Approximately 90.8% of long-form Census respondents were successfully linked to the CVSD.⁴⁶ Response and linkage rates differed by characteristics such as age, marital status, race/ethnicity, education, and income.⁴⁶ Statistics Canada provides sample weights to account for resulting sampling bias.

The population of interest for this study included adults aged 18 to 64 years who were active in the labor force during the week prior to Census Day (ie, employed or unemployed and actively seeking work). We excluded respondents younger than 18 or older than 64 years (n = 2 083 210) and those who were out of the labor force and not actively seeking work (eg, due to disability, caregiving, early retirement; n = 753 505). We also excluded those lacking information on our exposure of interest (ie, employment quality), which included respondents who were absent from work during the week prior to Census Day (n = 106 160) and respondents with no record of employment in 2005 (n = 122 375). We additionally excluded those with missing covariate information (n = 540). The final sample included 2 805 550 CanCHEC respondents. A study flow diagram is available in Appendix S1.

Measures

The exposure was employment quality at baseline (in 2006) and comprised 3 indicators: employment hours, employment stability, and employment earnings. These indicators are drawn from an established framework for assessing employment quality that has been widely applied in population health research and captures 3 of the dimensions within this framework: “working time arrangements,” “contractual stability,” and “material rewards.”^29-32,48

Employment hours refer to the number of hours worked during the preceding week. We grouped respondents into 4 categories: unemployed and looking for work (ie, 0 hours worked); 1 to 14 hours; 15 to 29 hours; 30 to 48 hours; and 49 or more hours. Employment stability refers to the number of weeks employed during the preceding calendar year. We grouped respondents into 3 categories: 1 to 25 weeks (ie, less than half the year); 26 to 51 weeks (ie, at least half the year but not the full year); and 52 weeks (ie, the full year). Employment earnings refer to pretax personal income earned from wages, salaries, or tips during the preceding calendar year. We grouped respondents into 5 quintiles, ranging from 1 (low) to 5 (high).

The outcomes of interest were death due to SDAM, which we examined separately. The underlying cause of death was determined using the International Classification of Diseases, 10th Revision (ICD-10), codes. Suicide and drug poisoning deaths were defined using Statistics Canada guidelines.⁴⁹ Alcohol-attributable deaths were defined using the Canadian Institute for Health Information indicator for causes that are 100% attributable to alcohol use.⁵⁰ A complete list of ICD-10 codes is presented in Appendix S2. We also recorded deaths from all other causes to account for competing risks.

Additional covariates included sex/gender (male or female), age group (18-24, 25-34, 35-44, 45-54, or 55-64 years), race/ethnicity (White, Black, East/Southeast Asian, Latin American, Middle Eastern, South Asian, or other/multiple), immigrant status (born in Canada or born outside Canada), marital status (single, married/cohabitating, or separated/divorced/widowed), household size (1, 2, 3, or ≥4), region (Eastern Canada, Central Canada, Western Canada, or Northern Canada), rurality (rural or urban), and education (less than high school, high school diploma, postsecondary less than bachelor’s degree, bachelor’s degree or higher). Although the Census only measures sex, we recognize that this measure captures the influence of both sex and gender on labor market outcomes and their health-related consequences. Therefore, we interpret this measure as an indicator of “sex/gender.”

Statistical analysis

We first described the baseline sociodemographic characteristics of the study population. We then used latent class analysis (LCA) to identify unobserved “classes” of respondents with similar employment conditions at baseline. This created an empirical typology of employment quality, reflecting varying degrees and forms of labor market attachment. Latent class analysis is particularly useful for assessing employment quality as a “package deal” of interrelated dimensions rather than treating them in isolation.^31,32 For example, part-time employment is not necessarily precarious, but when combined with low earnings or job instability, it can become so. On the other hand, part-time employment may offer certain advantages (eg, less stress, more leisure time) when combined with high earnings and job stability. By capturing these combinations of employment conditions, LCA provides a more nuanced and comprehensive understanding of how various dimensions of employment quality interact to influence health outcomes, rather than simply considering each employment dimension separately.

The LCA was conducted using the polca package in R (R Foundation for Statistical Computing).⁵¹ We included all 3 indicators in the LCA (employment hours, employment stability, and employment earnings). To determine the number of classes, we estimated multiple LCA models enumerating a progressively larger number of latent groups. We selected the most appropriate class solution based on a combination of formal fit indices and substantive interpretation of the results.⁵² Fit indices included the Akaike information criterion, the Bayesian information criterion, and entropy values.^53,54 Upon selecting the most appropriate class solution, we used descriptive statistics to summarize the prevalence and key features of the latent employment quality groups.

Directly age-standardized mortality rates for each employment quality group were estimated using Stata’s “stdize” command, with the 5-year age distribution of the 2011 Canadian Census population as the reference. We then examined the association between the employment quality typology and SDAM stratified by sex/gender. Sex/gender–stratified Poisson regression models were used to estimate rate ratios (RRs) and corresponding 95% CIs describing the association between employment quality and SDAM in separate models.⁵⁵ To account for length of follow-up, we included log person-years as an offset.⁵⁵ Follow-up was censored in the event of a competing (ie, non-SDAM) death.⁵⁶ For each outcome, we estimated a single fully adjusted model in which posterior probabilities of membership in each latent employment quality group were entered into the regression equation as continuous variables alongside the full set of covariates.

We conducted several sensitivity analyses. First, we excluded respondents who died within the first 2 years of follow-up, to mitigate the impact of reverse causation (e.g., preexisting mental health–related problems influencing employment quality). Second, we excluded respondents who were unemployed during the preceding week, to determine the extent to which associations were influenced by unemployment (as opposed to precarious employment per se). Third, we excluded respondents younger than 25 years because they may not yet have obtained their highest qualification. Fourth, to address potential age-related differences in the effects of employment quality on SDAM, we conducted a stratified analysis by age, distinguishing between respondents aged 18 to 44 years and those aged 45 to 64 years. Finally, to assess whether the latent classes were more strongly associated with SDAM than individual indicators, we examined the associations of each of the 3 employment quality indicators (hours, stability, and earnings) with SDAM independently.

Descriptive and regression analyses were conducted using SAS Enterprise Guide, version 8.1 (SAS Institute Inc.). All analyses applied sampling weights provided by Statistics Canada to ensure the generalizability of the results to the eligible Canadian Census population.

Results

Description of the study sample

The weighted study sample consisted of 7 113 445 women (46.9%) and 8 067 890 men (53.1%). A total of 51 155 SDAM-related deaths were observed in this weighted sample between 2006 and 2019. Of these deaths, 24 355 were related to suicide, 12 805 were related to drug poisonings, and 13 995 were related to alcohol-attributable causes.

The sociodemographic characteristics of the study population are described by sex/gender in Table 1. Among women and men, respectively, most identified as White (82.0% and 82.1%), were born in Canada (78.6% and 78.0%), were married or cohabiting (62.7% and 65.7%), lived in households with 4 or more people (38.9% and 41.4%), resided in Central Canada (62.2% 62.0%), resided in urban areas (81.7% and 80.2%), and attained postsecondary education less than a bachelor’s degree (38.4% and 37.8%).

Table 1.

Sociodemographic characteristics of the study population by sex/gender: Canadian Census Health and Environment Cohort (2006-2019).

	Women	Men
Overall no.	7 113 445	8 067 890
Age group, years, %
18-24	14.2	13.4
25-34	20.4	21.1
35-44	26.1	25.6
45-54	26.8	25.7
55-64	12.4	14.2
Race/ethnicity, %
White	82.0	82.1
Black	2.3	2.1
East/Southeast Asian	6.7	5.9
Latin American	1.0	1.0
Middle Eastern	0.9	1.3
South Asian	3.5	4.1
Other/multiple	3.6	3.4
Immigration status, %
Born in Canada	78.6	78.0
Born outside Canada	21.4	22.0
Marital status, %
Single	25.6	27.5
Married or cohabiting	62.7	65.7
Divorced, separated, or widowed	11.7	6.7
Household size, %
1	9.9	10.8
2	29.7	26.4
3	21.4	21.4
≥4	38.9	41.4
Region, %
Eastern Canada	7.1	6.9
Central Canada	62.2	62.0
Western Canada	30.4	30.8
Northern Canada	0.3	0.3
Rurality, %
Rural	18.3	19.8
Urban	81.7	80.2
Education, %
Less than high school	9.9	14.3
High school diploma	27.5	26.3
Postsecondary, less than bachelor’s degree	38.4	37.8
Bachelor’s degree or higher	24.1	21.6

Employment quality typology

The LCA models exploring a progressively larger number of latent classes indicated that the optimal solution fell between 3 and 6 classes. Model solutions with more than 5 classes did not result in substantially improved fit statistics or unique classes. The LCA model fit statistics for classes 2- to 8-class solutions are in Appendix S3. The 5-class solution was selected as the employment quality typology of choice because it yielded a diverse and substantively meaningful set of groups with varying degrees and forms of labor market attachment.

The employment characteristics of the 5 latent employment quality groups are presented in Table 2. The 5 employment quality groups are labeled “standard employment” (44.0%), “portfolio employment” (14.9%), “marginal employment” (13.2%), “intermittent employment” (16.3%), and “precarious employment” (11.6%). Standard employment, the largest of the groups, was characterized by a high probability of full-time hours (97.7% worked 30-48 h/wk), stable year-round employment (85.1% worked the full year), and gainful earnings (67.4% in the top 2 income quintiles). This group approximated the standard employment relationship, exhibiting favorable overall employment conditions that could be considered secure and rewarding. Portfolio employment was also characterized by favorable employment conditions, including relatively stable employment (76.3% worked the full year) and the highest earning levels (42.7% in the top income quintile). In contrast to the standard employment group, portfolio workers were likely to work long hours (98.9% worked ≥49 h/wk). This combination of features suggested engagement in employment that was secure but demanding.

Table 2.

Employment characteristics of the latent employment quality groups: Canadian Census Health and Environment Cohort (2006-2019).

	Standard	Portfolio	Marginal	Intermittent	Precarious
Overall, %	44.0	14.9	13.2	16.3	11.6
Employment hours, %
1-14	0.0	1.1	14.5	0.0	16.4
15-29	1.5	0.0	52.5	0.0	15.4
30-48	97.7	0.0	33.0	68.9	33.1
≥49	0.0	98.9	0.0	14.8	5.9
Unemployed and looking for worka	0.8	0.0	0.0	16.3	29.2
Employment stability, weeks, %
1-25 (less than half year)	0.0	0.3	0.0	6.2	86.8
25-51 (less than full year)	14.9	23.4	51.1	90.2	11.2
52 (full year)	85.1	76.3	48.9	3.6	2.0
Employment earnings quintile, %
1 (lowest)	0.0	12.3	52.8	0.0	70.2
2	11.4	8.5	30.0	46.4	27.9
3	21.2	11.9	12.5	48.7	1.1
4	36.2	24.7	3.6	3.4	0.4
5 (highest)	31.2	42.7	1.1	1.5	0.4

^aThe “unemployed” category for employment hours reflects unemployment in the week prior to the Census and no hours reported in that week.

Situated at the other end of the employment quality spectrum, precarious employment was characterized by a low probability of standard working hours (33.1% worked 30-48 h/wk), a high degree of instability (86.8% worked less than half the year), and the lowest earnings (70.2% in the lowest income quintile). This group also reported the highest rates of current unemployment (29.2%). With characteristics that starkly contrasted with the standard employment relationship, employment conditions in this group could be summarized as insecure and unrewarding.

We also identified 2 groups with intermediate employment quality. Marginal employment was characterized by a high probability of part-time hours (67.0% worked 1-29 h/wk) and low earnings (82.8% in the lowest 2 income quintiles), often accompanied by evidence of a job separation during the previous year (51.1% worked less than the full year). In summary, this group exhibited features of underemployment that are indicative of marginal labour market attachment; namely, limited hours and limited earnings.

Last, intermittent employment was characterized by a high probability of full-time hours (83.7% working ≥30 h/wk), albeit with evidence of substantial employment interruptions (96.4% worked less than the full year and 16.3% were currently unemployed) and modest earnings (95.1% in income quintiles 2 and 3). This group exhibited features of employment in seasonal and irregular jobs that are sporadic and unstable.

Sociodemographic characteristics of the latent employment quality groups are described by sex/gender in Table 3. Relative to those in standard or portfolio employment, respondents in lower-quality (ie, marginal, intermittent, and precarious) employment were more likely to be younger, single, non-White, and born outside of Canada. They were also less likely to have a bachelor’s degree or higher.

Table 3.

Sociodemographic characteristics of the latent employment quality groups by sex/gender: Canadian Census Health and Environment Cohort (2006-2019).

	Women					Men
	Standard	Portfolio	Marginal	Intermittent	Precarious	Standard	Portfolio	Marginal	Intermittent	Precarious
Overall, %	43.3	8.0	18.3	16.7	13.7	44.5	21.0	8.7	16.0	9.8
Age group, years, %
18-24	5.6	4.5	21.8	15.2	36.2	5.8	4.2	29.2	18.8	45.1
25-34	20.0	20.2	16.4	25.5	21.2	21.1	19.9	17.9	26.4	17.5
35-44	29.4	29.7	22.9	25.3	18.8	28.5	31.0	16.8	22.6	13.9
45-54	32.4	32.6	23.2	23.7	14.4	29.9	30.4	17.8	20.2	12.6
55-64	12.6	13.0	15.6	10.4	9.3	14.6	14.6	18.3	12.1	11.0
Race/ethnicity, %
White	84.5	83.7	83.2	77.9	76.2	84.2	86.6	77.6	77.7	74.5
Black	2.1	2.1	2.1	2.7	3.0	1.8	1.4	3.0	2.5	3.0
East/Southeast Asian	6.0	6.8	6.0	8.5	7.8	5.7	4.0	7.6	6.9	7.7
Latin American	0.7	0.7	1.1	1.3	1.4	0.9	0.7	1.3	1.5	1.3
Middle Eastern	0.6	0.8	1.0	1.1	1.5	1.0	1.1	2.1	1.7	2.2
South Asian	2.9	2.8	3.1	4.6	5.0	3.6	3.5	4.5	5.5	5.3
Other/multiple	3.1	3.1	3.5	4.0	5.1	2.8	2.6	4.0	4.1	6.0
Immigration status, %
Born in Canada	80.3	78.2	79.4	74.8	77.0	78.9	79.6	75.3	75.2	77.6
Born outside Canada	19.7	21.8	20.6	25.2	23.0	21.1	20.4	24.7	24.8	22.4
Marital status, %
Single	19.4	19.7	28.4	27.1	42.8	20.0	14.1	45.5	36.5	60.1
Married or cohabiting	67.3	66.3	62.6	60.6	49.1	72.9	79.2	48.2	56.7	34.5
Divorced, separated, or  widowed	13.3	14.0	9.0	12.3	8.1	7.1	6.7	6.3	6.8	5.4
Household size, %
1	11.6	13.7	6.6	10.1	6.4	11.1	9.8	10.4	12.4	9.6
2	32.1	32.6	27.0	31.4	22.4	27.4	25.0	27.0	27.5	21.8
3	21.1	19.0	21.1	22.3	23.4	21.5	19.8	21.4	22.2	23.1
≥4	35.2	34.7	45.3	36.3	47.8	40.0	45.3	41.2	37.9	45.5
Region, %
Eastern Canada	6.7	5.8	6.4	7.0	10.1	5.8	6.0	5.8	8.3	11.9
Central Canada	64.5	57.8	61.0	61.7	59.7	65.1	55.9	64.1	61.0	60.3
Western Canada	28.5	36.0	32.4	31.1	29.8	28.7	37.8	29.9	30.4	27.3
Northern Canada	0.4	0.4	0.2	0.2	0.4	0.3	0.3	0.2	0.3	0.5
Rurality, %
Rural	16.0	20.7	20.8	17.7	21.2	16.4	24.5	18.3	22.0	22.9
Urban	84.0	79.3	79.2	82.3	78.8	83.6	75.5	81.7	78.0	77.1
Education, %
Less than high school	7.1	7.9	12.5	12.2	13.6	11.3	13.0	17.4	19.0	20.3
High school diploma	24.4	20.5	32.3	28.1	34.2	23.7	22.6	32.9	29.0	35.7
Postsecondary, less  bachelor’s degree	40.5	35.3	37.4	40.9	32.1	40.9	38.6	31.8	37.4	28.0
Bachelor’s degree or  higher	27.9	36.3	17.8	18.8	20.1	24.2	25.9	18.0	14.6	16.0

Associations between employment quality and SDAM

Covariate-adjusted associations between the employment quality typology and SDAM by sex/gender are presented in Table 4, where the number of deaths, person-years, and age-standardized mortality rates are also described in the weighted sample by employment quality group.

Table 4.

Associations between employment quality and suicide, drug poisoning, and alcohol-attributable mortality by sex/gender: Canadian Census Health and Environment Cohort (2006-2019).

	Deaths	Person-years	ASMR (95% CI)	RR a  (95% CI)
Suicide mortality
Women
Standard	1 640	41 572 610	4.7 (4.4-4.9)	Ref.
Portfolio	385	7 703 005	5.7 (5.1-6.3)	1.48 (1.27-1.72)
Marginal	1 085	17 527 650	8.4 (7.9-9.0)	1.96 (1.75-2.19)
Intermittent	845	16 064 725	7.1 (6.6-7.6)	1.72 (1.48-1.99)
Precarious	805	13 124 495	8.7 (8.0-9.4)	2.43 (2.20-2.68)
Men
Standard	7 645	48 171 360	20.9 (20.3-21.5)	Ref.
Portfolio	3 815	22 704 885	21.4 (20.6-22.3)	1.08 (1.02-1.14)
Marginal	1 960	9 346 075	29.6 (28.2-31.1)	1.52 (1.41-1.63)
Intermittent	3 840	17 264 725	30.2 (29.1-31.2)	1.75 (1.64-1.87)
Precarious	2 335	10 605 195	34.3 (32.7-36.0)	1.69 (1.60-1.79)
Drug poisoning mortality
Women
Standard	1 155	41 572 610	3.4 (3.2-3.6)	Ref.
Portfolio	275	7 703 005	4.6 (3.8-5.4)	1.68 (1.40-2.01)
Marginal	810	17 527 650	6.4 (5.9-6.9)	2.05 (1.80-2.34)
Intermittent	810	16 064 725	6.4 (5.9-6.9)	3.10 (2.65-3.63)
Precarious	830	13 124 490	9.3 (8.6-10.1)	3.58 (3.21-4.00)
Men
Standard	2 840	48 171 360	8.7 (8.3-9.1)	Ref.
Portfolio	1 380	22 704 880	10.5 (9.6-11.3)	1.16 (1.07-1.27)
Marginal	1 030	9 346 075	15.8 (14.7-16.8)	1.84 (1.66-2.05)
Intermittent	1 990	17 264 725	14.7 (14.0-15.4)	2.41 (2.19-2.66)
Precarious	1 690	10 605 200	20.6 (19.4-21.8)	2.69 (2.49-2.90)
Alcohol-attributable mortality
Women
Standard	1 500	41 572 615	4.5 (4.2-4.7)	Ref.
Portfolio	245	7 703 005	4.2 (3.6-4.9)	0.99 (0.83-1.18)
Marginal	505	17 527 650	4.1 (3.7-4.5)	0.93 (0.81-1.08)
Intermittent	550	16 064 725	5.3 (4.8-5.8)	1.20 (1.01-1.44)
Precarious	440	13 124 495	6.1 (5.5-6.8)	1.77 (1.57-1.99)
Men
Standard	4 530	48 171 360	12.0 (11.6-12.4)	Ref.
Portfolio	1 920	22 704 880	10.1 (9.6-10.6)	0.90 (0.84-0.97)
Marginal	1 140	9 346 075	19.3 (18.1-20.4)	1.51 (1.37-1.66)
Intermittent	1 865	17 264 725	16.9 (16.0-17.7)	1.48 (1.35-1.62)
Precarious	1 300	10 605 195	26.4 (24.8-27.9)	2.22 (2.07-2.38)

Abbreviations: ASMR, age-standardized mortality rate per 10 000 people; Ref., referent; RR, rate ratio.

^aAdjusted for age, race/ethnicity, immigrant status, marital status, region, rurality, education, and household size.

We observed graded associations between employment quality and rates of SDAM. Across all 3 outcomes, age-standardized mortality rates were generally lowest among individuals in standard employment and highest among individuals in precarious employment, with those in marginal, intermittent, and, to a lesser extent, portfolio employment occupying intermediate positions. Absolute rate differences between employment quality groups tended to be larger among men than women, with the former having considerably higher baseline rates of SDAM.

Rate ratios, on the other hand, were generally larger among women in the case of suicide and drug poisoning deaths and among men in the case of alcohol-attributable deaths. Compared with standard employment, precarious employment was associated with 2.43 (95% CI, 2.20-2.68) greater rate of suicide, 3.58 (95% CI, 3.21-4.00) greater rate of drug poisoning, and 1.77 (95% CI, 1.57-1.99) greater rate of alcohol-attributable deaths among women; and a 1.69 (95% CI, 1.60-1.79), 2.69 (95% CI, 2.49-2.90), and 2.22 (95% CI, 2.07-2.38) greater rate, respectively, among men. With few exceptions, marginal and intermittent employment were also associated with elevated rates of SDAM, albeit not as high as precarious employment. Compared with those in standard employment, women in marginal and intermittent employment were substantially more likely to die by suicide (RR = 1.96 [95% CI, 1.75-2.19]; RR = 1.72 [95% CI, 1.48-1.99]) and drug poisoning (RR = 2.05 [95% CI, 1.80-2.34]; RR = 3.10 [95% CI, 2.65-3.63]), respectively; whereas men in marginal and intermittent employment were substantially more likely to die by all 3 outcomes, including suicide (RR = 1.52 [95% CI, 1.41-1.63]; RR = 1.75 [95% CI, 1.64-1.79]), drug poisoning (RR = 1.84 [95% CI, 1.66-2.05]; RR = 2.41 [95% CI, 2.19-2.66]), and alcohol-attributable causes (RR = 1.51 [95% CI, 1.36-1.66]; RR = 1.48 [95% CI, 1.35-1.62]), respectively. Of note, among women and men, intermittent employment had a stronger association with drug poisoning than did marginal employment—a pattern that did not extend to the other 2 outcomes.

Finally, portfolio employment was associated with elevated rates of suicide (RR = 1.48; 95% CI, 1.27-1.72) and drug poisoning (RR = 1.68; 95% CI, 1.40-2.01) deaths but not alcohol-attributable deaths among women, and, to a lesser extent, was associated with drug poisoning deaths (RR =1.16; 95% CI, 1.07-1.27) but not suicide or alcohol-attributable deaths among men.

Several sensitivity analyses supported the robustness of our primary analyses. Excluding respondents who died during the first 2 years of follow-up (Appendix S4) or those who were unemployed during the preceding week (Appendix S5) revealed no substantial deviations from our main findings. Excluding respondents younger than 25 years produced nearly identical findings, with the exception that employment quality was more strongly associated with drug poisoning deaths among men in this age-restricted sample (Appendix S6). Stratification by age showed that, in some instances, relative associations were stronger among older respondents (aged 4-64 years) compared with younger respondents (aged 18-44 years) (Appendix S7). This was particularly true for precarious employment, which was more strongly associated with SDAM among older workers. When considering baseline age-standardized rates among younger and older respondents, absolute differences for alcohol-attributable deaths were larger among older workers, whereas absolute differences for drug poisoning deaths tended larger among younger workers for men. Estimates were otherwise comparable across younger and older workers. Finally, the latent classes were more strongly associated with SDAM than were the individual employment quality indicators of hours, stability, and earnings when considered separately (Appendix S8).

Discussion

This study examined associations between employment quality and death due to SDAM causes in the Canadian working population. Using latent class clustering methods to derive a typology of employment quality, we identified 5 distinct groups reflecting varying degrees and forms of labor market attachment, labeled (generally in order of decreasing employment quality) standard, portfolio, marginal, intermittent, and precarious. We found consistent evidence of a mortality gradient across employment quality groups, with lower-quality employment in general—and precarious employment in particular—being associated with increased rates of SDAM compared with those in higher-quality (ie, standard) employment. Relative associations between employment quality and SDAM were uniquely patterned by sex/gender. Whereas most estimates in the analysis of suicide and drug poisoning deaths were stronger among women, most estimates in the analysis of alcohol-attributable deaths were stronger among men. Taken together, our findings highlight important and nuanced associations among employment quality, SDAM, and sex/gender in a population of economically active workers in Canada.

The employment quality typology we identified reflects the heterogeneous nature and structure of employment relations after the deregulation of the labor market and subsequent erosion of the standard employment relationship in Canada from the 1990s onward.^57-59 Our typology aligns in key respects with parallel studies conducted in the United States and Europe.^31-33 This international literature has characterized both the standard employment relationship as well as various other forms of employment (eg, portfolio, precarious) that deviate from the standard ideal type in terms of hours, stability, and earnings. Previous Canadian studies examining employment quality in relation to disability⁶⁰ and sexual orientation⁶¹ derived employment quality typologies using similar methods and identified comparable standard, portfolio, and precarious employment types. These and the aforementioned international studies identified additional employment types (eg, instrumental, inflexible, dead-end) that were not identified in our LCA. This may reflect (1) the set of employment-related variables available in the Census relative to the social surveys used in prior studies, which tend to provide a more diverse array of indicators; (2) the historical nature of our data, which reflect labor market conditions in 2006; and (3) cross-national differences in labor market and employment contexts. Nevertheless, we identified intermediate marginal and intermittent employment types that, in some ways, approximate the groups we did not identify (eg, nonstandard hours combined with low to moderate stability and pay). Overall, the 5 identified employment quality groups form a continuum of employment conditions, ranging from standard employment (ie, secure and rewarding) at 1 end to precarious employment (ie, insecure and unrewarding) at the other.^62,63

Our findings on the associations between employment quality and SDAM align with existing literature emphasizing the significant health risks associated with low-quality employment.^62,63 We demonstrate that individuals in lower-quality employment experience greater rates of SDAM compared with those in higher-quality standard employment. Thus, when it comes to understanding the labor market drivers of SDAM, it appears necessary to address both employment status (ie, does the person have a job?) and employment quality (ie, does the person have a good-quality job) as key risk factors. These insights from our study align with previous research from the United States, Sweden, Australia, Belgium, South Korea, and various other high-income countries describing an association between lower-quality (eg, precarious) employment and higher rates of depression, suicidal ideation, psychological distress, substance use disorders, and other mental health–related outcomes.^33-42,64-66 Our findings extend the literature by demonstrating a further link to a crucial downstream end point, namely, premature death due to SDAM.

The results of our study also highlight important sex/gender differences in the association between employment quality and SDAM. Across all 3 outcomes, we observed larger absolute inequalities among men—a pattern that most likely reflects their larger baseline rates of SDAM.^67,68 However, relative inequalities were generally larger among women in the case of suicide and drug poisoning deaths and larger among men in the case of alcohol-attributable deaths. Previous literature has noted how sex/gendered roles and social norms (eg, the prevailing assumption that men should be the primary “breadwinner” of their household) can shape the employment-health relationship, with considerable evidence of a stronger relationship among men, particularly when it comes to the health effects of unemployment.^69-71 In contrast, women face greater pressure to balance multiple and competing responsibilities at work and home.^69-71 Low-quality employment may aggravate the stress associated with this gendered division of labor and, in particular, the double burden of paid and unpaid work that women disproportionately bear.^71-74 Furthermore, other gendered stressors—such as sexual and gender discrimination in the workplace—may be more prevalent in precarious employment.^61,75 These dynamics, in turn, could explain why employment quality was more strongly associated with suicide and drug poisoning deaths among women relative to men. We also found precarious employment was more strongly associated with SDAM among older than younger workers, consistent with the idea that precarious employment at later ages may be a marker of cumulative disadvantage with greater consequences for health.^76,77 Evidence of heterogeneity across sex/gender and age groups represents an important area for future research that may provide crucial insights for preventive interventions.

Our study findings have important policy implications because they provide evidence that underscores the importance of creating and promoting high-quality employment for supporting mental health and well-being in the working population. These include jobs that offer adequate hours, stability, and earnings to fulfill the financial and psychological needs of workers. Public health policies should be designed and implemented to promote and maintain high-quality employment conditions. This is particularly important given the rising rates of SDAM, the declining quality of employment, and the observed connection between precarious employment conditions and risk of SDAM. Taken together, our findings suggest employment quality may represent a promising upstream and modifiable target for policy and programmatic interventions aiming to reduce the burden of SDAM.

Strengths of our study include the use of the 2006 CanCHEC, which links data from respondents to the mandatory long-form Census respondents with 13 years of follow-up data in national mortality records. Data from this large nationally representative cohort enabled the precise evaluation of associations between employment quality and SDAM by sex/gender. Furthermore, by assessing employment quality using a multidimensional typology, our study offers an advantage over simplistic assessments of employment quality based on single indicators (eg, temporary vs permanent, part-time vs full-time) and emphasizes the significance of multiple, overlapping dimensions of employment.^78,79

Despite these important strengths, our study should be interpreted considering the following limitations. Our employment quality typology was constructed using data collected in 2006 (ie, nearly 20 years ago). Although this allowed for sufficient follow-up, employment conditions have changed substantially over the past 2 decades, particularly with the emergence of a digital gig economy. Additionally, employment quality was assessed using 3 indicators reflecting employment hours, stability, and earnings. Other dimensions of employment quality, such as rights and social protection and interpersonal power relations, were not measured in the 2006 Census and, therefore, are missing from our operationalization of the exposure.^29-32 Although the LCA suggested a suitable degree of separation between classes, future studies may define employment groups more discretely by considering additional indicators of employment quality (eg, union membership, benefit coverage). Moreover, the use of a single-point-in-time measure of employment quality may contribute to an underestimation of its association with SDAM. However, previous studies have found substantial evidence of stability in employment quality over an individual’s working life course.^33,36 Last, data on preexisting mental health risk factors were not collected in the Census, limiting the ability to examine reverse causation, whereby underlying mental health conditions can influence employment quality and risk of SDAM. The absence of rich health information about CanCHEC participants may also limit our ability to account for a range of potential confounding factors influencing both employment quality and risk of SDAM. Future longitudinal studies linking a range of life-course health data may provide additional insight into causal relationships between employment quality and SDAM.

We found that lower-quality employment is associated with increased mortality rates due to SDAM causes—so-called deaths of despair. Policy interventions that target and improve employment conditions may help reduce the burden of premature deaths attributable to suicide and substance use. Future research should explore the mechanisms linking employment quality and SDAM, investigate the heterogeneous nature of the relationship by sex/gender, age, and across specific “despair”-related outcomes, and quantify the potential impact of improvements in employment quality on population health and health equity.

Supplementary material

Supplementary material is available at the American Journal of Epidemiology online.

Acknowledgments

The authors thank Amélie Quesnel-Vallée (Department of Epidemiology, Biostatistics and Occupational Health, Department of Sociology, McGill University) for providing helpful comments on the study protocol.

Funding

This research did not receive any direct funding from agencies in the public, private, or not-for-profit sectors. The Institute for Work and Health is supported through funding from the Ontario Ministry of Labour, Immigration, Straining, and Skills Development (MLITSD).

Conflict of interest

The authors declare no conflicts of interest.

Disclaimer

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

Data availability

CanCHEC data were accessed through the Statistics Canada Research Data Centre Program, which is subject to strict confidentiality and disclosure protocols indicated in the Statistics Act of Canada. These data can be accessed upon request and approval from Statistics Canada (see process here: https://www.statcan.gc.ca/eng/rdc/index). All research outputs were vetted by Statistics Canada before release to ensure privacy.

Ethical standards

Data for this study were accessed through the Statistics Canada Research Data Centre Program, which is subject to strict confidentiality and disclosure protocols indicated in the Statistics Act of Canada. These data are exempt from formal ethics review and approval, as stipulated in Article 2.2(a) of the Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans. All residents of Canada were legally required to complete the 2006 Census, according to the Statistics Act, and all participants in this study provided implicit consent for use of their data.