Article Text
PDF
Abstract
Background Occupying a low socioeconomic position is associated with increased mortality risk. To disentangle this association, previous studies considered various dimensions of socioeconomic trajectories across the life course. However, they used a limited number of stages. We simultaneously examined various dimensions of the whole professional trajectory and its association with mortality.
Methods We used a large sample (337 706 men and 275 378 women) of the data obtained by linking individuals’ annual occupation (collected in 1976–2002 from a representative panel of the French salaried population in the semipublic and private sectors) with causes of death obtained from registries. All-cause and cause-specific HRs were estimated using Cox's regression models adjusted for the occupational class at the beginning of the follow-up, the current occupational class, the transition rates between occupational categories and the duration of time spent in occupational categories.
Results An increase in the time spent in the clerk class increased men and women's cardiovascular mortality risk compared with that in the upper class (HRs=1.59 (1.14 to 2.20) and 2.65 (1.14 to 6.13) for 10 years increase, respectively, for men and women). Men with a high rate of transitions had about a 1.2-fold increased risk of all-cause and external-cause mortality compared with those without transitions during their professional life. This association was also observed for women's all-cause mortality.
Conclusions Strong associations between professional trajectories and mortality from different causes of death were found. Long exposure to lower socioeconomic conditions was associated with increased mortality risk from various causes of death. The results also suggest gradual associations between transition rates and mortality.
- MORTALITY
- SOCIAL INEQUALITIES
- SOCIAL CLASS
- Life course epidemiology
Statistics from Altmetric.com
Introduction
Socioeconomic inequalities in mortality, as quantified by mortality differentials between social groups, have been studied in many industrialised countries.1–3 Despite the low level in mortality and its continuous decrease, studies conducted in the UK, USA and Europe have shown that these inequalities are still large in some countries4–6 and have increased over time in men as well as women.1 ,7–11
A large body of research has shown that mortality rates are higher among those in lower socioeconomic positions12 ,13 regardless of the socioeconomic indicator (occupational status, educational level or income).14 Most of these studies have measured socioeconomic positions only at one stage of life. This approach does not consider the impact of transitions between different socioeconomic groups. Thus, to obtain better understanding of the relationship between health and socioeconomic position, various dimensions of socioeconomic trajectories, such as the evolution of socioeconomic position and the modality of transitions between social groups, need to be taken into account.15 ,16
Although an observed individual's social level at a given time reflects his/her social position at different stages of his/her past life,17 several life-course models have been proposed in the literature to explain the possible association between socioeconomic status and health: critical period, accumulation and social mobility models. The critical period model considers some stages or specific moments in life as key periods affecting health. The cumulative model hypothesises that mortality differentials are explained by the accumulation of all present and past working conditions, lifestyles and behaviours. Analyses using this model are based on the life-cumulated length of stay in the most disadvantaged social group. They suggest that the accumulation of poor socioeconomic exposure in life increases the risk of mortality.15 ,18–20 The social mobility model was developed to take into account the modality of transitions between social groups once or several times in life. These models help to explain the potential impact of socioeconomic status on health. However, some studies point to a bias in the results due to the impact of health on socioeconomic position. Therefore, this reverse causation is another issue that should be taken into account.21 ,22
The aim of this study was to examine the associations between life-course professional trajectory and adult mortality. Previous studies have considered two or three stages in professional life and used a simple classification for socioeconomic positions (low-medium-high). This paper goes further by considering the whole professional trajectories as well as all-cause and cause-specific mortality. For this purpose, we use life-course models on a representative sample of the French salaried population in the semipublic and private sectors from 1976 to 2002 to investigate the possible ways in which professional trajectories may be associated with adult mortality.
Methods
Cosmop-DADS database
The Cosmop-DADS database was obtained by linking the occupational life course provided from the panel of the Annual Declarations of Social Data (DADS),23 which has been regularly updated by the French National Institute for Statistics and Economic Studies (INSEE) since 1976, with the causes of death recorded by the French National Death Registry (INSERM-CépiDc). The DADS Panel contains the employment records of approximately 1/24th of all employees in the private and semipublic sectors, that is, 80% of all paid occupations in France. Episodes of careers declared as self-employed, employees of the state, employees in agriculture, domestic services, extra-territorial activities, interns and apprentices are excluded from its scope. A deterministic record linkage using the following identifiers linked these two data sets: sex, date of birth, date of death and the commune of residence at the time of death. The matching rate was 98%. In total, the Cosmop-DADS population is a sample of the French population (comprising of people for whom vital status and date of death are available), employed at least once as salaried workers in the semipublic and private sectors between 1976 and 2002.
The study was approved by the French data protection committee and institutional ethical review board: Commission Nationale de l'Informatique et des Libertés (CNIL; authorisation n° 904210v1).
Occupational classes
Occupations were coded according to the French classification created by INSEE regarding various social characteristics, without any specific hierarchical order between the defined classes.24 Originally, the DADS covered five classes: craftsmen and trade-related workers; upper class; intermediary occupations; clerk class and manual workers class.
Since DADS declarations are mandatory for employers, there were theoretically no missing occupational episodes for employees working in companies within the DADS scope. However, professional trajectories were not fully observed for several individuals. The first set of missing episodes were for the years 1981, 1983 and 1990, which were not collected owing to administrative reasons. These episodes were complemented with information from the previous years. For the other years, some occupations could not be classified in the five occupational classes. These occupations were imputed using a multivariable multinomial logistic regression25 incorporating sex, age and type of employment in the imputation model.
Since regional and local authorities were not fully covered by DADS declarations before 1987, any occupations of this type were excluded from our professional scope. The same decision was taken for occupations declared in the craftsmen and trade-related workers class, as those in DADS are not representative of this class in the general population.
In summary, our professional scope contained the DADS scope mentioned previously excluding regional and local authorities, and craftsmen and trade-related workers class. Those outside this scope could either be working, inactive or retired. It was not possible to distinguish these different situations. As it is well established that inactivity is associated with an increased mortality risk,26 ,27 any episodes outside this scope should not be ignored, so the category ‘outside the scope’ was added to the four other categories. This strategy induced a bias, but given the structure of the data, building a sound imputation model would require additional assumptions for which no auxiliary data, such as data on employees of the public sector, were available.
Study population
All individuals born in the French territories for whom a salaried period was declared in Cosmop-DADS between ages 25 and 30, excluding those working outside the study scope in their first year (337 706 men and 275 378 women) were included in the study (owing to the uncertainty of the vital status of people born outside France, they were excluded from our study population). Less than 1% of occupations were imputed, and in total, 22% and 30% of follow-up years were outside the study scope for men and women, respectively. Fifty-two percent of men and 61% of women were outside the study scope for at least 1 year of their follow-up. Owing to the non-negligible number of episodes outside the study scope and the lack of available information for making more hypotheses about these episodes, a replicated analysis was carried out on a subsample of the analysed population for whom the first 5 years of their follow-up was covered by the study scope in order to ensure that an observed trajectory was complete (in the first 5 years) for the analysis (198 381 men and 134 784 women, with fewer than 14% of follow-up years outside the study scope in total).
Professional trajectory
A professional trajectory may be defined as the sequence of consecutive professional positions occupied by an individual (figure 1). To characterise it, three time-dependent variables were used:
-
Occupational class at each year;
-
Cumulative social class indicator, defined as individual's length of stay in each occupational class. This indicator was calculated for all classes except the upper class, so the latter served as reference;
-
Ten-year social mobility indicator, defined by the transition rates between classes, excluding the ‘outside the scope’ category and calculated as follows:
Examples of fictional trajectories. Example: Individual 2 was working in the manual workers class from 1978 until 1985. He was outside the scope of the study between 1986 and 2001 and finally worked in an intermediary occupation in 2002.
This indicator was categorised into three groups using tertiles, separately for men and women.
To limit the impact of reverse causation,21 ,22 occupational classes were considered with a 2-year time lag, that is, instead of using the current occupational class, that of 2 years before was taken into account.
Mortality
The Cosmop-DADS database is a sample of the French population for whom the vital status and date of death are available. All individuals of this sample were followed up to 2002 and the administrative censoring date was set at 31 December 2002. The underlying causes of death, recorded by INSERM-CépiDc, were coded according to the International Classification of Diseases, 8th, 9th and 10th revisions (ICD-8, ICD-9 and ICD-10). Three broad categories of causes were specifically considered: cardiovascular diseases, cancer and external causes (see online supplementary appendix I).
Statistical analysis
Cox proportional hazards models were used to estimate all-cause HRs, cause-specific HRs (CSHs) and their 95% CIs while accounting for left truncation induced by the delayed entries. Age was used as the time scale.28 The model for each cause was fitted using a Cox model by censoring the participants who failed from competing cause.29
Adjustment for the variables, occupational class at the beginning of the follow-up as a baseline covariate and the three indicators of professional trajectory as time-dependent covariates, was achieved by performing univariable analysis in the first step and then using all these covariates in a multivariable analysis. Considering the decrease in mortality rates over time in France, the models were adjusted for observation periods.
The occupational class and social mobility indicator were introduced into the models as categorical variables, and the upper class and those without any mobility between classes were considered as the reference categories. For the cumulative social class indicator, HRs were interpreted as the hazard corresponding to an increase in the time spent in an occupational class versus that spent in the upper class. These HRs were calculated for a 10-year increase. No violation of the proportional hazards assumptions was found according to Schoenfeld residuals.
Proportional hazards models were conducted separately for men and women using the Survival package of R software,30 and the imputation was carried out by IVEware software.31
Results
The average number of transitions between occupational classes differed between the age categories. Transitions were more numerous between the ages of 25 and 44 in women, and between the ages of 25 and 34 in men. At the beginning of the follow-up, the largest class was the clerk class (about 54%) in women and manual workers (about 60%) in men. For young men (25–34 years), 49.3% of the cumulated time spent was in the manual workers class and much less in the upper class (6.5%). The same magnitude was observed in young women for the clerk and upper class (25–34 years; table 1).
Characteristics of study population according to occupational trajectories
During the follow-up, 12 162 (3.6%) men and 3551 (1.3%) women died. Most deaths occurred between the ages of 35 and 44. In total, 48.7% of deaths among women and 39.8% of deaths among men occurred while individuals were outside the study scope 2 years before death. Most other deaths in men and women occurred while they were in the manual workers class and the clerk class, respectively (table 2).
Distribution of study population according to occupational trajectories
Overall, the same magnitude was found for the results of the univariable and multivariable analysis, except for the estimated HRs for the social mobility indicator, although, adjusting for all indicators led to some attenuation in the increased risk of death in association with professional trajectory indicators. The results of the multivariable analysis are subsequently presented (those of the univariable analysis could be found in online supplementary appendix II).
Occupation at beginning of follow-up
Men in the manual workers class at the beginning had a higher mortality risk compared with those who were in the upper class (except for cancer mortality) but to a different degree depending on the causes of death (table 3). In women, this association was not statistically significant (table 4).
All-cause and cause-specific mortality HRs among men according to socioprofessional trajectories
All-cause and cause-specific mortality HRs among women according to socioprofessional trajectories
Current occupational class
Among men, being in the clerk class increased the mortality risk compared with being in the upper class (HRs=1.49 (1.31 to 1.69), 1.58 (1.09 to 2.30), 1.50 (1.16 to 1.93), 1.43 (1.14 to 1.79) and 1.58 (1.26 to 1.98), respectively, for mortality from all causes, cardiovascular diseases, cancer, external causes and other causes). Among men, those in the manual workers class had an increased mortality risk compared with those in the upper class (HRs=1.39 (1.25 to 1.56), 1.43 (1.03 to 1.99), 1.26 (1.02 to 1.56) and 1.73 (1.42 to 2.12), respectively, for all-cause, cardiovascular, cancer and external-cause mortality). Those outside the study scope had the highest mortality risk except for cardiovascular and cancer mortality among women, that is, about twofold to threefold higher than the mortality risk in the upper class (tables 3 and 4).
Cumulative time spent in occupational classes
The cumulative time spent in occupational classes was strongly associated with men's all-cause and cause-specific mortality and women's all-cause and cardiovascular mortality, with less pronounced associations for men's external-cause mortality. Among men, more time spent in an occupational class increased the mortality risk compared with that in the upper class. This increase in manual workers was associated with a 1.8-fold higher cancer mortality risk (HR=1.75 (1.48 to 2.06)) and that outside the study scope was associated with a 1.5-fold higher external-cause mortality risk (HR=1.46 (1.19 to 1.77)) compared with that in the upper class. Among women, more time spent in the clerk class was associated with a 2.7-fold higher cardiovascular mortality risk compared with that in the upper class (HR=2.65 (1.14 to 6.13); tables 3 and 4).
Social mobility indicator
In the univariable analysis, an inverse association between the social mobility indicator and mortality was systematically found among men, and only for cancer mortality among women. Adjusting for other indicators changed the direction of the results, except for women's cancer mortality.
In multivariable analysis, the same magnitude was observed for this indicator among men and women except for women's external-cause mortality, with significant results for men and women's all-cause, external-cause and other causes mortality, and women's cancer mortality. Having a high social mobility indicator increased the all-cause mortality risk (HRs=1.15 (1.09 to 1.21) and 1.13 (1.04 to 1.22), respectively, for men and women), the other causes mortality risk (HRs=1.23 (1.12 to 1.34) and 1.40 (1.19 to 1.64), respectively, for men and women) and the external-cause mortality risk (HR=1.17 (1.08 to 1.28) for men) compared with not experiencing any mobility during professional life (tables 3 and 4).
Ad hoc sensitivity analysis
When replicated analyses were performed on the subsample, including individuals working in the study scope during their first 5 years of follow-up, the estimated all-cause and CSHs did not change for any of the indicators except for men's cardiovascular mortality (see online supplementary appendix III).
Discussion
Previous studies on this topic have generally considered individuals’ socioeconomic position at two or three stages of life including childhood (father's socioeconomic position), entry into the labour market and mid-life position. To our knowledge, the present study is the first to investigate the association between the whole professional trajectory and all-cause mortality and within that, three major causes of death: cardiovascular disease, cancer and external causes. Overall, our results add to the existing evidence of the strong relationship between professional trajectory and all-cause mortality among men, with less pronounced associations among women.13 ,15 ,16 ,32–36
Compared with previous studies, a new aspect of our study is the use of the duration of time spent in occupational classes as a measure of socioeconomic exposure and the transition rates between occupational classes as a measure for capturing the social mobility dimension.
The three most commonly used life-course models, namely the critical period, cumulative and social mobility models were taken into account. Our results suggest that all three dimensions are associated with men's all-cause mortality. For women, only the cumulative and the social mobility models were confirmed by this analysis.
Interpretations and comparisons with other studies
As shown in previous studies, strong associations between professional trajectories and men's and women's mortality was found.13 ,15 ,16 ,32–36 However, a direct comparison with other studies cannot be easily made given the different occupational classifications in each country, and the fact that we used whole professional trajectories.
The present study only focused on professional trajectories with no information on childhood circumstances. However, the individual's first occupation is likely to be the most representative dimension of the end of childhood. We found that the association between the first occupation and mortality was strong for men's cardiovascular and external-cause mortality. Previously, strong associations have also been reported between socioeconomic circumstances in childhood and mortality from some causes of death, such as cardiovascular diseases.15 ,32 ,33 ,35
On the other hand, for some other causes of death such as external causes and lung cancer,35 stronger associations were found between socioeconomic circumstances in adulthood and adult mortality than those in childhood. Our results are in accordance with the literature, since in other studies, for some causes of death such as external causes and cancer, occupational classes were found to be strongly linked with men's mortality. Supplementary analysis on different cancers also reported the same associations or even stronger ones (for deaths by UADT cancers; data not shown). For women, the results were not statistically significant.
Another hypothesis in the literature is the putative association between the accumulation of exposure to different socioeconomic conditions and mortality. However, the use of only three stages of life limited the number of possible trajectories, so the different trajectories could be compared. By investigating the duration of time spent in each occupational class instead of comparing different trajectories, we found a strong relationship between the duration of exposure to low professional position and mortality. This association was stronger for cardiovascular and cancer mortality in men but was significant only for all-cause and cardiovascular mortality in women. This is consistent with the results of previous studies.15 ,16 ,33 ,37 The large mortality risk of those who stay longer in the low occupational categories can be explained by exposure to poor working conditions and by the fact that the least skilled are less likely to move upward. Furthermore, staying a long time in the same professional conditions could reflect a greater adherence to a professional class and its specific lifestyle.
The changes between occupational categories and their dynamics were also pointed out in previous studies. Some studies have shown that within classes, male movers have a mortality risk situated between that of non-movers in their class of origin and that of their destination.38 ,39 We investigated the association between the frequency of changes between occupational classes and mortality. Instability in professional life may be interpreted in two ways. If instability is chosen, it could be the reflection of high dynamism with the ability to change and adapt to several professional environments. Conversely, if instability is forced, it could be due to difficulties in finding one's place, to a high dependence on the work market or to personal events. We found an inverse association for this indicator in the univariable analysis, as it does not take into account the occupational classes before and after the transitions. Our results of the multivariable analysis show that participants with high transition rates have an increased risk of all-cause and external-cause mortality. These results suggest that the instability measured is more forced than chosen, with a deleterious association on mortality. In a very explorative approach to disentangle the chosen and forced instability, we considered the following naive order of occupations from high to low level: ‘upper class’, ‘intermediary occupations’, ‘clerk class’ and ‘manual workers’. Although this order is not strictly hierarchical, upward and downward changes were studied as separate variables. The risk of mortality was positively associated with downward changes (eg, going from the ‘upper class’ to the ‘clerk class’) and negatively with upward changes (eg, going from the ‘manual workers class’ to the ‘intermediary occupations class’; data not shown).
Limitations
The main limitation in this investigation is the high percentage of follow-up years outside the scope of the study. The decision to consider all these data in the ‘outside the scope’ category could induce a bias. However, we examined a wide range of occupational sectors and the occupational stages are sufficiently reliable as they were collected within the context of administrative procedures. Furthermore, the replicated analysis on the subsample with sufficient follow-up provided almost the same results, which strengthens the findings.
All participants had worked at least once between the ages of 25 and 30 and were likely to be healthier than the general population, so the sample should not be interpreted as representative of the French population.
Finally, taking into account the individual's occupation with a 2-year time lag could reduce the reverse causation bias. However, for some causes of death such as transport accidents, the problem of reverse causation is less likely to be a source of bias.
Despite these drawbacks, the large size of the sample, the annual nature of the information collected and the causes of death coded with high precision are the major strength of this study. Using repeated measures of occupational category over the follow-up could provide insight into changes that may have occurred during a person's professional life. To gain a better understanding of the complex social inequalities in mortality, future analysis should focus on models that take into account, simultaneously, all aspects of professional trajectories and mortality. Joint modelling of nominal occupational data and cause-specific mortality following the approach of Li et al40 is the object of an ongoing project.
What is already known on this subject?
-
Previous studies reported strong associations between socioeconomic trajectories and mortality.
-
Most of these studies have used two or three stages of life to show these associations across life-course models.
What this study adds?
-
We consider all stages of professional trajectories to investigate these relationships in a representative sample of the French salaried population from semipublic and private sectors.
-
Long-time exposure to poor socioeconomic position was strongly associated with adult mortality, especially for cardiovascular diseases.
-
Having more transitions during professional life was adversely associated with mortality.
Acknowledgments
The authors would like to thank Walid Ghosn for his help and comments during this study.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Footnotes
-
Contributors MK contributed to the analysis and interpretation of data and drafted the manuscript. AL contributed to the statistical design of the study. BG-P and AF contributed to the interpretation of data and revised the manuscript. GR designed and supervised the study.
-
Funding This work was supported by joint funding from the Ministry of Health, the General Directorate of Health and the Mission research of the Directorate of research, studies, evaluation and statistics, the National Health Insurance Fund of the Salaried Workers, the Social Security Scheme for Self-employed Workers, the National Solidarity Fund for Autonomy and the National Institute of Prevention and Education for Health, within the framework of the 2011 call for research launched by IReSP (Institute of Research in Public Health) (grant number A11226LS).
-
Competing interests None.
-
Ethics approval This study was approved by the French data protection committee and institutional ethical review board: Commission Nationale de l'Informatique et des Libertés (authorisation n° 904210v1).
-
Provenance and peer review Not commissioned; externally peer reviewed.