Skip to main content
Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Association between unemployment and the co-occurrence and clustering of common risky health behaviors: Findings from the Constances cohort

  • Marie Plessz ,

    Roles Conceptualization, Data curation, Formal analysis, Methodology, Project administration, Writing – original draft, Writing – review & editing

    Affiliations Centre Maurice Halbwachs (EHESS, ENS, CNRS, PSL), INRAE, Paris, France, Centre Maurice Halbwachs (EHESS, ENS, CNRS, PSL), Paris, France

  • Sehar Ezdi,

    Roles Writing – review & editing

    Affiliation Centre Maurice Halbwachs (EHESS, ENS, CNRS, PSL), Paris, France

  • Guillaume Airagnes,

    Roles Conceptualization, Writing – review & editing

    Affiliations Department of Psychiatry and Addictology, Hôpitaux Universitaires Paris Ouest, Paris, France, UMS 011, Population-based Epidemiological Cohorts, Inserm, Villejuif, France

  • Isabelle Parizot,

    Roles Conceptualization, Writing – review & editing

    Affiliation Centre Maurice Halbwachs (EHESS, ENS, CNRS, PSL), Paris, France

  • Céline Ribet,

    Roles Project administration, Resources, Writing – review & editing

    Affiliation UMS 011, Population-based Epidemiological Cohorts, Inserm, Villejuif, France

  • Marcel Goldberg,

    Roles Resources, Writing – review & editing

    Affiliation UMS 011, Population-based Epidemiological Cohorts, Inserm, Villejuif, France

  • Marie Zins,

    Roles Methodology, Resources, Writing – review & editing

    Affiliation UMS 011, Population-based Epidemiological Cohorts, Inserm, Villejuif, France

  • Pierre Meneton

    Roles Conceptualization, Formal analysis, Methodology, Writing – original draft, Writing – review & editing

    Affiliation UMR 1142 LIMICS, Inserm, Sorbonne Université, Université Paris 13, Paris, France



Unemployment is associated with a high prevalence of risky health behaviors. Mortality increases with the number of co-occurring risky behaviors but whether these behaviors co-occur with a greater than expected frequency (clustering) among unemployed people is not known.


Differences according to unemployment status in co-occurrence and clustering of smoking, alcohol abuse, low leisure-time physical activity and unhealthy diet (marked by low fruit and vegetable intake) were assessed in 65,630 salaried workers, aged 18 to 65, who were participants in Constances, a French population-based cohort. Among them, 4573 (7.0%) were unemployed without (n = 3160, 4.8%) or with (n = 1413, 2.1%) past experience of unemployment.


Compared to the employed, unemployed participants without or with past experience of unemployment were similarly overexposed to each risky behavior (sex and age adjusted odds-ratios ranging from 1.38 to 2.19) except for low physical activity, resulting in higher rates of co-occurrence of two, three and four behaviors (relative risk ratios, RRR 1.20 to 3.74). Association between behavior co-occurrence and unemployment did not vary across gender, partnership status or income category. Risky behavior clustering, i.e., higher than expected co-occurrence rates based on the prevalence of each behavior, was similar across unemployment status. The same observations can be made in employed participants with past experience of unemployment, although overexposure to risky behaviors (ORs 1.15 to 1.38) and increased rates of co-occurrence (ORs 1.19 to 1.58) were not as pronounced as in the unemployed.


Co-occurrence of risky behaviors in currently and/or formerly unemployed workers is not worsened by behavior clustering. Engagement in each of these behaviors should be considered an engagement in distinct social practices, with consequences for preventive policies.


Rising unemployment rates over the last decade in most European and North American countries [1] has attracted growing attention on the public health impact of job loss [2]. Indeed, unemployment is thought to raise premature mortality [3] by increasing the incidence of suicide [4], poor mental health [5], cancer [6, 7] and cardiovascular disease [812]. The mechanisms by which unemployment would increase the incidence of these pathologies remain elusive but overexposure to behavioral risk factors is likely to be involved [13, 14]. By far, the leading behavioral causes of premature mortality in Western populations are alcohol abuse, smoking, unbalanced diet and low physical activity [15].

Alcohol misuse was first suspected to be linked to unemployment during the industrial revolution in the 19th century [16]. Since then, the evidence suggesting that unemployed people are at increased risk of heavy alcohol intake, binge drinking and/or alcohol use disorders has accumulated. A review of the literature between 1990 and 2010 reports several studies showing that unemployment increases alcohol use and the incidence of alcohol disorders [16]. For example, in middle-aged Americans, becoming unemployed raises the risk of developing alcohol abuse/dependence six-fold compared to those who remain in employment [17]. More recently, positive and significant associations have been described between job loss during the past year and average daily ethanol consumption, number of binge drinking days and the probability of alcohol abuse and/or dependence diagnosis in large samples of the American population [18, 19]. In the Northern Swedish Cohort Study, non-moderate alcohol consumption in middle-aged adults has been associated with a higher exposure to unemployment during their youth [20]. Likewise, data from the Christchurch Health and Development Study in New Zealand showed that unemployment of at least three months’ duration significantly increases the risk of alcohol use disorder in young adults [21].

A higher risk of smoking and/or increased frequency of tobacco use is another unhealthy behavior that has been convincingly linked to job loss and unemployment [16]. For example, unemployed middle-aged Americans consume more cigarettes per day if they already smoke and have a greater risk of relapse if they are ex-smokers in comparison to those who remain employed [22]. Unemployed young Americans who smoke are also less likely to attempt cessation than the employed [23]. More recent studies have confirmed that unemployment is associated with a higher risk of smoking in large samples of the American population [18, 24, 25]. This research has been corroborated by similar findings from the Scottish and German populations [26, 27].

Several studies suggest that unemployment may also lead to unhealthy food habits. For example, a long history of unemployment in Finnish young adults has been shown to be a good predictor of stress-related eating characterized by a high consumption of sausages, hamburgers, pizzas and chocolate [28]. Unemployment has been associated with low consumption of starchy foods, fruits and vegetables, seafood and dairy products in a deprived middle-age French population attending food aid organizations [29]. Being unemployed has also been linked to food insecurity measured by the necessity of buying cheaper food and/or low consumption of fruits and vegetables in New Zealand [30]. Likewise, job insecurity/unemployment in Portuguese adults has been associated with an unhealthy dietary pattern characterized by low consumption of soups, vegetables, fresh fruits, fish, dairy products and high meat consumption [31]. Other studies have reported aggregated data showing that increased unemployment rates are associated with reduced consumption of fruits and vegetables and increased consumption of unhealthy foods such as snacks and fast food in North American populations [32, 33]. In Danish households, different dietary behaviors are observed depending on duration of unemployment, i.e., a higher intake of fat and protein due to increased consumption of animal-based foods immediately after job loss and a higher intake of carbohydrates and added sugar thereafter [34].

Low physical activity may be another unhealthy consequence of unemployment. Although physical activity at work varies substantially from one occupation to another [35], being unemployed has been associated with a reduction in daily physical activity in American adults [36] but not in Swedish adults [37]. Even leisure-time activity is modified among the unemployed. It has been shown that unemployed Swedish adults have lower leisure-time physical activity compared to the employed although the difference disappears when adjusting for education level [38]. Similarly, American adults who have been unemployed for a year or more [39], as well as young unemployed Americans [40], have less leisure-time physical activity than those who are employed. Compared to their employed counterparts, unemployed Finnish adults also more frequently report economic constraints and the lack of companionship as barriers for leisure-time physical activity [41].

The co-occurrence of behavioral risk factors dramatically increases the effects on health and mortality. Thus, old European men and women who combine adherence to a Mediterranean diet, moderate alcohol use, being physically active and non-smoking have a mortality rate one third of that which is observed in those who do not adopt any of these behaviors [42]. Similarly, middle-aged and older UK men and women who combine non-smoking, being physically active, moderate alcohol intake and fruit and vegetable intake of at least five servings a day have a 4-fold difference in the risk of dying over an average period of 11 years compared to those who do not adhere to any of these behaviors [43]. In addition to the mere co-occurrence of behavioral risk factors, there can exist a clustering of these factors, i.e., a higher frequency of co-occurrence than expected on the basis of the prevalence of each factor, that can exacerbate the effects on health and mortality [44]. To our knowledge, despite the large body of evidence showing that unemployment is associated with a high exposure to common behavioral risk factors, no study has examined the co-occurrence and clustering of these factors among unemployed people. One report has documented relative prevalence rates of cigarette smoking, risky drinking, non-engagement in leisure-time physical activity and low fruit/vegetable consumption among unemployed young US adults but has not explored the co-occurrence or clustering of these risky health behaviors [40]. This is the purpose of the present study, which investigates the co-occurrence and clustering of common unhealthy behaviors in French adults who are unemployed and/or have been unemployed in the past. The potential implication for health policies is to determine whether unemployed people need specific preventive strategies targeting reciprocal relationships between unhealthy behaviors, as would be the case if unemployment is associated with a clustering of these behaviors.

Materials and methods

Study population

The CONSTANCES cohort includes adults selected between 2012 and 2018 from the French population covered by the general health insurance system (over 85% of the population) according to a random sampling scheme stratified on age, sex, socioeconomic status and region [45]. The 15% excluded were mostly farmers or self-employed workers who had never worked as salaried workers. Inclusion criteria comprised of written informed consent, a comprehensive health examination in one of the 21 participating medical centers scattered across French metropolitan territory, and self-administered questionnaires on lifestyle, health-related behaviors, social and occupational conditions at inclusion and in the past (inclusion rate was 7.3%). The study received approval from both the Ethics Evaluation Committee of the French National Institute of Health and Medical Research (Inserm) and the National Committee for the Protection of Privacy and Civil Liberties (Cnil).

In the cohort at the time of data extraction, there were 91,259 adults aged 18 to 65, currently or previously salaried workers, who declared that they were currently either employed or unemployed and for whom lifestyle, occupational exposure and professional schedule questionnaires were completed. We excluded individuals who declared that they were not working for health reasons (149 unemployed) and who had missing data on covariates (12,172) or dependent variables (13,308). This resulted in a study population of 65,630 participants. Among them, 51,875 (79.1%) were never unemployed, 9,182 (14%) were unemployed in the past but not at inclusion, 3,160 (4.8%) were unemployed at inclusion but not in the past and 1,413 (2.1%) were unemployed both in the past and at inclusion.

Exposure variable: Employment status

The employment status of participants at inclusion was assessed by a question with multiple choices, allowing participants to describe complex situations. Possible answers were: “I have a job (even if on sick leave, unpaid leave or availability, maternity, paternity, adoption or parental leave)”, “Unemployed or job seeker”, “Retired or no longer in business”, “In training (pupil, student, trainee, apprentice, etc.)”, “I do not work for health reasons (long-term illness, disability)”, “No professional activity”. Participants who ticked the box “I have a job”, and only this one, were considered employed. Participants who ticked the box “Unemployed or job seeker” were considered unemployed only if they had not also ticked the boxes “I have a job” or “I do not work for health reasons”. Lifetime unemployment was documented by a separate questionnaire in which participants were asked to report each time they had stopped working for a period of more than six months, and why (unemployment, health, other). By combining these data with those at inclusion, four types of experience of unemployment were defined: participants who were unemployed at inclusion and at least once in the past, those who reported being unemployed at least once in the past but not at inclusion, those who were unemployed at inclusion but reported no prior unemployment, and those who had a job and reported no earlier unemployment spell longer than 6 months. For the sake of clarity, we refer to the two last groups as “unemployed at inclusion only” and ‘never unemployed’ but it should be remembered that these participants might have had short periods of unemployment in the past.

Outcome variables: Co-occurring behavioral risk factors

The main outcome variable is the number of behavioral risk factors each participant was exposed to. The four risk factors considered were collected in a self-completed questionnaire at inclusion. Consumption of fruits and vegetables was determined through a food frequency questionnaire covering a regular week and used as a proxy for diet quality [46]. Data from this questionnaire have already been published [47]. People were considered at risk if the sum of their fruit (fresh or squeezed) and vegetable (raw or cooked) consumption was lower than three times per day, everyday (not just the day before the questionnaire). This cut-off is consistent with the definitions of low fruit and vegetable (FV) consumption in French health policy [48].

Leisure-time physical activity was determined by a calculated score ranging from 0 (i.e. being very active) to 6 (not being active at all). The physical activity questionnaire asked about regular practice of walking or cycling; practicing a sport; gardening or housekeeping over the past 12 months. Each of the three items was noted 0 if the answer was no; 1 if practice was regular but low (less than 15 minutes for sports, or 2 hours for the other two weekly); 2 if practice was regular and higher. Data regarding the score obtained by summing the three items has been published [49]. People with a score below three were considered at risk of low physical activity.

Smoking was understood as participants who smoked at inclusion, excluding consumption of e-cigarettes or cannabis. Alcohol abuse referred to drinking habits during the week before completing the questionnaire and was defined as a consumption exceeding two or three drinks per day in women and men respectively.


Age was divided in three categories (18–36, 37–47 and 48–65 years). Educational attainment was classified into three levels: university, secondary school, primary school. Income comprised monthly earnings of all household members and was collected in seven categories chosen according to the distribution of household disposable income in France in 2013 [50]. It was recoded as low (below 1500 euros), middle (between 1500 and 2800 euros) or high (above 2800 euros). A ´low´ household income assigned the participant to the first quintile of household income, while a ‘high’ household income assigned the participant above the median. Whether the household was single-headed and had children was also documented, as well as the region of residence grouped into six geographical areas (Paris area, north-east, south-west, south-east, Brittany, center area). As an indicator of overall physical and psychological health status of participants, self-rated health was assessed with an eight-level scale that was reduced to three levels for the analyses: good (levels 1 and 2), average (levels 3 and 4) or poor (levels 5 to 8 roughly corresponding to the 90th percentile).

Statistical analyses

First, we computed descriptive statistics, including prevalence of each behavioral risk factor, by experience of unemployment.

Second, we analyzed co-occurrence. We defined co-occurrence as the number of risk factors the participants were exposed to. We estimated the association between risk co-occurrence at inclusion and experience of unemployment using multinomial logistic regressions. The base level of the outcome variable was “to be exposed to zero risk” and the reference category was “never unemployed”. The models yielded relative risk ratios (RRR) for each number of co-occurring risk and each category of unemployment, as well as 95% confidence intervals and statistical tests for trends across unemployment category. Three models were applied: in M1 we minimally adjusted for sex and age; in M2 we added education, single-adult household, household with children, region of residence and self-rated health; in M3 we also added household monthly income category. We also tested interactions between experience of unemployment and each covariate in order to check whether some groups of participants were more at risk of co-occurring behavioral risk factors.

Third, we examined risk clustering. Clustering of risk factors refers to the co-occurrences of risk factors with greater frequencies than expected by chance, i.e. if exposure to each risk were independent of one other [44]. For each number of co-occurring risk factors, expected rates of co-occurrence were computed from the prevalence of each risk factor assuming that they occurred independently. Clustering was defined as observed to expected prevalence ratios significantly >1. We also computed the frequency and clustering of each specific risk combinations. All the analyses were performed with the Stata software (version 15, Stata Corp., College Station, TX).


Characteristics of participants at inclusion according to their experience of unemployment

As reported in Table 1, the characteristics of participants who were unemployed at inclusion without or with past experience of unemployment were very similar except for age, as those who were unemployed in the past were obviously more likely to be older. Compared to participants who never encountered unemployment, they were less educated and belonged more often to single-adult households without children and with low monthly income. Their geographical distribution was slightly different and they declared more often poor or average self-rated health.

Table 1. Characteristics of participants at inclusion according to their experience of unemployment.

Unemployed participants reported more often low FV intake. The minimally adjusted OR (95% CI) was 1.38 (1.25–1.53) for the unemployed at inclusion without past experience of unemployment. The unemployed were also more exposed to smoking (2.06 (1.91–2.23)) and alcohol abuse (1.78 (1.60–1.96)), but not low leisure-time physical activity (1.02 (0.94–1.10), p = 0.60, and 1.10 (0.97–1.23–1.31), p = 0.11). For the unemployed with past experience of unemployment, the OR were slightly but not significantly larger than for those without past experience of unemployment. Except for age, the characteristics of participants who were unemployed in the past but not at inclusion were generally intermediary, between those of participants who never encountered unemployment and those of participants who were unemployed at inclusion without or with past experience of unemployment. This also applied to low FV intake (1.14 (1.08–1.21)), smoking (1.38 (1.31–1.45)) and alcohol abuse (1.22 (1.13–1.32)). Exposure to low leisure-time physical activity was highest among those who were unemployed in the past but not at inclusion (1.20 (1.14–1.25)).

Co-occurrence of risky health behaviors in participants at inclusion according to their experience of unemployment

Compared to participants who never encountered unemployment, those who were unemployed at inclusion without or with past experience of unemployment were similarly likely to be exposed to one risky behavior rather than none (minimally adjusted RRR 1.17 and 1.11) and more likely to be exposed to two (RRR 1.65 and 1.82), three (RRR 2.51 and 2.47) or four behaviors (RRR 2.95 and 3.74). This is shown in Table 2, model 1. There was a gradient in the association between exposure to unemployment and exposure to two, three or four risky behaviors, as shown by the highly significant p for trends.

Table 2. Relative risk ratios of co-occurring risk factors according to experience of unemployment in Constances cohort: Multinomial regression (reference: 0 risk).

As a result, the frequency ranking of co-occurrence of risky behaviors differed somewhat according to unemployment status (Table 3). For those exposed in the past and at inclusion, the most frequent was to be exposed to two behaviors (37.5%) followed by one (35.3%), three (14.4%) and zero (10%). For those never exposed, the most frequent was one behavior (46%), followed by two (30.5%) and zero (13.4%) with only 1.1% exposed to four behaviors.

Table 3. Clustering of behavioral risk factors in participants at inclusion according to their experience of unemployment (observed-to-expected ratios).

When covariates were added to the minimally adjusted models, the RRR decreased but remained statistically significant for two, three or four risky behaviors. This was also true when income category was included. For example, for the unemployed with past experience of unemployment the RRR for two, three or four behaviors were 1.28, 1.39 and 1.78 respectively (Table 2, model 3).

When we examined interactions between covariates and experience of unemployment in the fully adjusted model, there were statistically significant interactions for age (chi-square test p = 0.015), self-rated health (p = 0.052) and education (p = 0.036). However, when we stratified models over these variables, there was no pattern or trend in the RRR (S2 Table). Association between risky behavior co-occurrence and unemployment did not vary across gender, partnership status, or income category.

Risky health behavior clustering according to experience of unemployment

Table 3 shows evidence of behavior clustering, whatever the experience of unemployment: the observed numbers of participants with none, three or four risky behaviors was greater than expected. The numbers of participants with one or two behaviors were smaller.

The lack of participants having one or two risky behaviors was similar in all experiences of unemployment. The excess of participants having no risky behavior (as compared to expected numbers) was higher when they were unemployed at inclusion without or with past experience of unemployment than when they never encountered unemployment. Conversely, the greater-than-expected exposure to three or four risky behaviors tended to be lower for participants having experienced unemployment of any kind, suggesting that having experienced unemployment was associated with smaller behavior clustering. However, the differences between O/E ratios across experience of unemployment only reached statistical significance for the exposure to four risky behaviors. In summary, risky behavior clustering was observed across all experiences of unemployment but was not stronger among those with past and/or present exposure of unemployment than in those who never experienced unemployment.

As shown in Tables 3 and 4, the three most frequent combinations of risky behaviors were low FV intake (30.9 to 40.3%), low FV intake and low physical activity (13.7 to 18.7%), and low FV intake and smoking (9.3 to 16.1%), with slight variations in the frequency order across experience of unemployment. The two combinations with the highest observed-to-expected ratios were all four behaviors (O/E 1.95 to 2.70), and FV & smoking & alcohol abuse (1.63 to 1.92). Beyond that, the order of risky behavior combinations varied across experience of unemployment.

Table 4. Combinations of behavioral risk factors in participants at inclusion according to their experience of unemployment.


The present study reports the occurrence, co-occurrence and clustering of common risky behaviors in adults who were unemployed at the time of the analyses and/or had encountered unemployment in the past.

Prevalence rates of exposure to none or only one risky behavior were at the higher end of the distributions described in other Western populations while rates of exposure to two, three or four behaviors were at the lower end [5159]. The most frequent co-occurrences of two risky behaviors were low FV intake associated with either smoking, leisure-time physical inactivity or alcohol abuse, in agreement with what has been reported in the English population [60]. The most common exposures to three risky behaviors, i.e., low FV intake and smoking associated with either leisure-time physical inactivity or alcohol abuse were also among those that have been the most frequently observed in other European populations [54, 57, 59]. Likewise, greater than expected proportions of people exposed to none, two, three or four risky behaviors have already been documented in several Western populations [53, 54, 57, 59, 61, 62].

Unemployment at inclusion (~7% rate, slightly lower than the current national average which is around 9% but very close to the mean European rate) was correlated with significant differences in the frequency of co-occurrence of risky behaviors. The first observation is that unemployment was associated with higher prevalence rates of each behavior, adding to the large body of evidence already discussed in the introduction. The difference was quite substantial for smoking and alcohol abuse (1.7-fold in both cases), in agreement with the literature which has reported average increases of 1.6 and 2.2-fold respectively in unemployed people [16]. In contrast, the difference was much more limited for low FV intake and low leisure-time physical activity (less than 10% in both cases). Low FV intake was very frequent, which could explain this small increase. It is more difficult to account for the small variation in leisure-time physical activity whose prevalence has been shown in some studies to be 1.6-fold higher in unemployed people [38, 39].

As a consequence of the higher prevalence of each risky behavior, unemployed participants were more frequently exposed to two, three or four behaviors, resulting in a further increase in their mortality risk [43]. This was true even after controlling for income category, indicating that income differences contribute partly but not entirely to the association between unemployment and risky behaviors. Stratified analyses indicated that among unemployed participants, there was no variation in behavior co-occurrence across gender, household status or income category. Clustering of risky behaviors was observed, in the sense that the observed prevalence of zero, three or four behaviors were higher than those expected if exposures had been statistically independent. However, clustering was similar across experience of unemployment, or even slightly weaker among the unemployed.

Essentially, similar results were obtained in unemployed participants without or with past experience of unemployment, suggesting that potential effects of current unemployment on the engagement in risky behaviors prevail over those of past unemployment. However, most of the observations can also be made in employed participants who have been unemployed in the past, although to a smaller extent than in unemployed participants. This supports the view that the increased prevalence and co-occurrence of risky behaviors would be a consequence of unemployment and that this might contribute to poor health in unemployed people [2, 3], although analyses on prospective data are required to support this assumption.

The present study has several limitations. First, the study selected only salaried or formerly salaried participants, excluding self-employed or farmers (who were not covered by the French unemployment insurance system at the time of the study). Second, the very low participation rate resulted in the selection of motivated and socially privileged individuals even though the stratified sampling strategy tried to compensate for the higher non-response of people with a low socioeconomic status. This is illustrated by a comparison of cohort participants with randomly selected workers in the same age range [63]. The proportion of participants with university education is very high and the unemployment rate is somewhat lower (S1 Table). Third, unemployment status and risky behaviors were self-reported. Reporting bias is possible and may vary across population categories. More specifically some experiences of unemployment were not captured by the questionnaire, such as unemployment upon labor-market entry (the questionnaire starts with the first job lasting at least 6 months) or short alternating episodes of unemployment and paid jobs (the questionnaire records episodes lasting at least six months). The number and actual duration of episodes of unemployment in the past or at inclusion were also not available for the analyses. Last, the cutoff points for exposure to risky behaviors, although chosen consistently with French public health guidelines, might affect the results.

In conclusion, this study shows that current unemployment, and to a lesser extent past unemployment, was associated with increased engagement in common risky health behaviors, supporting the view that these behaviors might partly mediate increased morbidity and mortality in unemployed people [3]. However, we found that risky behaviors did not cluster more among participants experiencing unemployment than among those in employment. This might be due to the fact that engagement in risky behaviors relies on mechanisms that differ from one behavior to another, hence, the prevalence of one behavior can increase independently of others, regardless of the employment status. Engagement in risky behaviors may be related to strategies for coping with stress but may also involve multiple understandings of when and how much it is appropriate to consume alcoholic drinks, cigarettes or healthy foods, or engage in any practice that generates physical activity [64]. This suggests that preventive strategies should be adapted to each unhealthy behavior among unemployed as well as employed people, and would benefit from a practice-based approach, thereby acknowledging that they constitute distinct social practices with specific meaning, context, and embeddedness in social relations [65].

Supporting information

S1 Table. Baseline sociodemographic characteristics of CONSTANCES participants compared to randomly selected workers.


S2 Table. Relative risk ratios of co-occurring risk factors according to experience of unemployment in Constances cohort: Stratification by age, education and self-rated health, multinomial regression (reference: 0 risk).



  1. 1. OECD.Stat (2018) Short-Term Labour Market Statistics: Harmonised Unemployment Rates (HURs).
  2. 2. Brand JE (2015) The Far-Reaching Impact of Job Loss and Unemployment. Annu Rev Sociol 41: 359–375. pmid:26336327
  3. 3. Roelfs DJ, Shor E, Davidson KW, Schwartz JE (2011) Losing life and livelihood: a systematic review and meta-analysis of unemployment and all-cause mortality. Soc Sci Med 72: 840–854. pmid:21330027
  4. 4. Milner A, Page A, LaMontagne AD (2013) Long-term unemployment and suicide: a systematic review and meta-analysis. PLoS One 8: e51333. pmid:23341881
  5. 5. Paul KI, Moser K (2009) Unemployment impairs mental health: Meta-analyses. Journal of Vocational Behavior 74: 264–282.
  6. 6. Antunes JL (2016) The impact of unemployment on cancer mortality, and how to avoid it. Ann Transl Med 4: 404. pmid:27867956
  7. 7. Lynge E (1997) Unemployment and cancer: a literature review. IARC Sci Publ 343–351. pmid:9353675
  8. 8. Dupre ME, George LK, Liu G, Peterson ED (2012) The cumulative effect of unemployment on risks for acute myocardial infarction. Arch Intern Med 172: 1731–1737. pmid:23401888
  9. 9. Gallo WT (2012) Evolution of research on the effect of unemployment on acute myocardial infarction risk. Arch Intern Med 172: 1737–1738. pmid:23165943
  10. 10. Mejean C, Droomers M, van der Schouw YT, Sluijs I, Czernichow S, Grobbee DE et al. (2013) The contribution of diet and lifestyle to socioeconomic inequalities in cardiovascular morbidity and mortality. Int J Cardiol 168: 5190–5195. pmid:23998549
  11. 11. Meneton P, Kesse-Guyot E, Méjean C, Fezeu L, Galan P, Hercberg S et al. (2015) Unemployment is associated with high cardiovascular event rate and increased all-cause mortality in middle-aged socially privileged individuals. Int Arch Occup Environ Health 88: 707–716. pmid:25385250
  12. 12. Weber A, Lehnert G (1997) Unemployment and cardiovascular diseases: a causal relationship? Int Arch Occup Environ Health 70: 153–160. pmid:9298397
  13. 13. Janlert U (1997) Unemployment as a disease and diseases of the unemployed. Scand J Work Environ Health 23 Suppl 3: 79–83.
  14. 14. Jin RL, Shah CP, Svoboda TJ (1995) The impact of unemployment on health: a review of the evidence. CMAJ 153: 529–540. pmid:7641151
  15. 15. Mokdad AH, Marks JS, Stroup DF, Gerberding JL (2004) Actual causes of death in the United States, 2000. JAMA 291: 1238–1245. pmid:15010446
  16. 16. Henkel D (2011) Unemployment and substance use: a review of the literature (1990–2010). Curr Drug Abuse Rev 4: 4–27. pmid:21466502
  17. 17. Catalano R, Dooley D, Wilson G, Hough R (1993) Job loss and alcohol abuse: a test using data from the Epidemiologic Catchment Area project. J Health Soc Behav 34: 215–225. pmid:7989666
  18. 18. Compton WM, Gfroerer J, Conway KP, Finger MS (2014) Unemployment and substance outcomes in the United States 2002–2010. Drug Alcohol Depend 142: 350–353. pmid:25042761
  19. 19. Popovici I, French MT (2013) Does Unemployment Lead to Greater Alcohol Consumption. Ind Relat (Berkeley) 52: 444–466.
  20. 20. Virtanen P, Lintonen T, Westerlund H, Nummi T, Janlert U, Hammarström A (2016) Unemployment in the teens and trajectories of alcohol consumption in adulthood. BMJ Open 6: e006430. pmid:27016242
  21. 21. Boden JM, Lee JO, Horwood LJ, Grest CV, McLeod GF (2017) Modelling possible causality in the associations between unemployment, cannabis use, and alcohol misuse. Soc Sci Med 175: 127–134. pmid:28088618
  22. 22. Falba T, Teng HM, Sindelar JL, Gallo WT (2005) The effect of involuntary job loss on smoking intensity and relapse. Addiction 100: 1330–1339. pmid:16128722
  23. 23. Weden MM, Astone NM, Bishai D (2006) Racial, ethnic, and gender differences in smoking cessation associated with employment and joblessness through young adulthood in the US. Soc Sci Med 62: 303–316. pmid:16029919
  24. 24. Arcaya M, Glymour MM, Christakis NA, Kawachi I, Subramanian SV (2014) Individual and spousal unemployment as predictors of smoking and drinking behavior. Soc Sci Med 110: 89–95. pmid:24727666
  25. 25. Golden SD, Perreira KM (2015) Losing jobs and lighting up: Employment experiences and smoking in the Great Recession. Soc Sci Med 138: 110–118. pmid:26079992
  26. 26. Marcus J (2014) Does Job Loss Make You Smoke and Gain Weight? Economica 81: 626–648.
  27. 27. Semple S (2015) Employment, smoking, and health: the role of the hygienist. Ann Occup Hyg 59: 529–533. pmid:25925743
  28. 28. Laitinen J, Ek E, Sovio U (2002) Stress-related eating and drinking behavior and body mass index and predictors of this behavior. Prev Med 34: 29–39. pmid:11749094
  29. 29. Mejean C, Deschamps V, Bellin-Lestienne C, Oleko A, Darmon N, Hercberg S et al. (2010) Associations of socioeconomic factors with inadequate dietary intake in food aid users in France (The ABENA study 2004–2005). Eur J Clin Nutr 64: 374–382. pmid:20087381
  30. 30. Carter KN, Lanumata T, Kruse K, Gorton D (2010) What are the determinants of food insecurity in New Zealand and does this differ for males and females. Aust N Z J Public Health 34: 602–608. pmid:21134063
  31. 31. Gregório MJ, Rodrigues AM, Eusébio M, Sousa RD, Dias S, André B et al. (2017) Dietary Patterns Characterized by High Meat Consumption Are Associated with Other Unhealthy Life Styles and Depression Symptoms. Front Nutr 4: 25. pmid:28660194
  32. 32. Dave DM, Kelly IR (2012) How does the business cycle affect eating habits. Soc Sci Med 74: 254–262. pmid:22137244
  33. 33. Milicic S, DeCicca P (2017) The Impact of Economic Conditions on Healthy Dietary Intake: Evidence from Fluctuations in Canadian Unemployment Rates. J Nutr Educ Behav 49: 632–638. pmid:28889852
  34. 34. Smed S, Tetens I, Bøker Lund T, Holm L, Ljungdalh Nielsen A (2017) The consequences of unemployment on diet composition and purchase behaviour: a longitudinal study from Denmark. Public Health Nutr 1–13.
  35. 35. Proper KI, Hildebrandt VH (2006) Physical activity among Dutch workers—differences between occupations. Prev Med 43: 42–45. pmid:16684560
  36. 36. Van Domelen DR, Koster A, Caserotti P, Brychta RJ, Chen KY, McClain JJ et al. (2011) Employment and physical activity in the U.S. Am J Prev Med 41: 136–145. pmid:21767720
  37. 37. Kwak L, Berrigan D, Van Domelen D, Sjöström M, Hagströmer M (2016) Examining differences in physical activity levels by employment status and/or job activity level: Gender-specific comparisons between the United States and Sweden. J Sci Med Sport 19: 482–487. pmid:26095371
  38. 38. Ali SM, Lindström M (2006) Psychosocial work conditions, unemployment, and leisure-time physical activity: a population-based study. Scand J Public Health 34: 209–216. pmid:16581714
  39. 39. Siegel PZ, Brackbill RM, Heath GW (1995) The epidemiology of walking for exercise: implications for promoting activity among sedentary groups. Am J Public Health 85: 706–710. pmid:7733433
  40. 40. Caban-Martinez AJ, Lee DJ, Goodman E, Davila EP, Fleming LE, LeBlanc WG et al. (2011) Health indicators among unemployed and employed young adults. J Occup Environ Med 53: 196–203. pmid:21270653
  41. 41. Borodulin K, Sipilä N, Rahkonen O, Leino-Arjas P, Kestilä L, Jousilahti P et al. (2016) Socio-demographic and behavioral variation in barriers to leisure-time physical activity. Scand J Public Health 44: 62–69. pmid:26392420
  42. 42. Knoops KT, de Groot LC, Kromhout D, Perrin AE, Moreiras-Varela O, Menotti A et al. (2004) Mediterranean diet, lifestyle factors, and 10-year mortality in elderly European men and women: the HALE project. JAMA 292: 1433–1439. pmid:15383513
  43. 43. Khaw KT, Wareham N, Bingham S, Welch A, Luben R, Day N (2008) Combined impact of health behaviours and mortality in men and women: the EPIC-Norfolk prospective population study. PLoS Med 5: e12. pmid:18184033
  44. 44. McAloney K, Graham H, Law C, Platt L (2013) A scoping review of statistical approaches to the analysis of multiple health-related behaviours. Prev Med 56: 365–371. pmid:23518213
  45. 45. Goldberg M, Carton M, Descatha A, Leclerc A, Roquelaure Y, Santin G et al. (2017) CONSTANCES: a general prospective population-based cohort for occupational and environmental epidemiology: cohort profile. Occup Environ Med 74: 66–71. pmid:27884936
  46. 46. Aljadani HM, Patterson A, Sibbritt D, Hutchesson MJ, Jensen ME, Collins CE (2013) Diet quality, measured by fruit and vegetable intake, predicts weight change in young women. J Obes 2013: 525161. pmid:24062946
  47. 47. Plessz M, Kesse-Guyot E, Zins M, Matta J, Czernichow S (2019) Poverty does not modify the association between perceived diet healthiness and adherence to nutritional guidelines in the Constances cohort (France). Appetite 138: 190–197. pmid:30928517
  48. 48. Fourcade N, von Lennep F, Bauchet E, Grémy I, Vernay M (2017) L’état de santé de la population en France. Ministère des Solidarités et de la Santé
  49. 49. Matta J, Hoertel N, Kesse-Guyot E, Plesz M, Wiernik E, Carette C et al. (2019) Diet and physical activity in the association between depression and metabolic syndrome: Constances study. J Affect Disord 244: 25–32. pmid:30296663
  50. 50. Institut national de la statistique et des études économiques (2016) Les revenus et le patrimoine des ménages. Insee Références
  51. 51. Buck D, Frosini F (2012) Clustering of unhealthy behaviours over time: Implications for policy and practice. Kings Fund London: 1–24.
  52. 52. Chiolero A, Wietlisbach V, Ruffieux C, Paccaud F, Cornuz J (2006) Clustering of risk behaviors with cigarette consumption: A population-based survey. Prev Med 42: 348–353. pmid:16504277
  53. 53. Falkstedt D, Möller J, Zeebari Z, Engström K (2016) Prevalence, co-occurrence, and clustering of health-risk behaviors among people with different socio-economic trajectories: A population-based study. Prev Med 93: 64–69. pmid:27663427
  54. 54. Poortinga W (2007) The prevalence and clustering of four major lifestyle risk factors in an English adult population. Prev Med 44: 124–128. pmid:17157369
  55. 55. Randell E, Pickles T, Simpson SA, Spanou C, McCambridge J, Hood K et al. (2015) Eligibility for interventions, co-occurrence and risk factors for unhealthy behaviours in patients consulting for routine primary care: results from the Pre-Empt study. BMC Fam Pract 16: 133. pmid:26453044
  56. 56. Sabia S, Nabi H, Kivimaki M, Shipley MJ, Marmot MG, Singh-Manoux A (2009) Health behaviors from early to late midlife as predictors of cognitive function: The Whitehall II study. Am J Epidemiol 170: 428–437. pmid:19574344
  57. 57. Schuit AJ, van Loon AJ, Tijhuis M, Ocké M (2002) Clustering of lifestyle risk factors in a general adult population. Prev Med 35: 219–224. pmid:12202063
  58. 58. Tsai J, Ford ES, Li C, Zhao G, Pearson WS, Balluz LS (2010) Multiple healthy behaviors and optimal self-rated health: findings from the 2007 Behavioral Risk Factor Surveillance System Survey. Prev Med 51: 268–274. pmid:20647019
  59. 59. Zwolinsky S, Raine G, Robertson S (2016) Prevalence, Co-Occurrence and Clustering of Lifestyle Risk Factors Among UK Men. Journal of Men’s Heath 12: 15–24.
  60. 60. Meader N, King K, Moe-Byrne T, Wright K, Graham H, Petticrew M et al. (2016) A systematic review on the clustering and co-occurrence of multiple risk behaviours. BMC Public Health 16: 657. pmid:27473458
  61. 61. Shankar A, McMunn A, Steptoe A (2010) Health-related behaviors in older adults relationships with socioeconomic status. Am J Prev Med 38: 39–46. pmid:20117555
  62. 62. Tobias M, Jackson G, Yeh LC, Huang K (2007) Do healthy and unhealthy behaviours cluster in New Zealand. Aust N Z J Public Health 31: 155–163. pmid:17461007
  63. 63. Lanoe JL, Makdessi-Raynaud Y (2005) L’état de santé en France en 2003: santé perçue, morbidité déclarée et recours aux soins à travers l’enquête décennale santé. Etudes et Résultats 436: 1–12.
  64. 64. Blue S, Shove E, Carmona C, Kelly MP (2016) Theories of practice and public health: understanding (un)healthy practices. Critical Public Health 26: 36–50.
  65. 65. Cohn S (2014) From health behaviours to health practices: an introduction. Sociology of Health & Illness 36: 157–162.