Job loss from poor health, smoking and obesity: a national prospective survey in France

Transcription

1 Evidene-based poliy and pratie 1 Institut de Reherhe et Doumentation en Eonomie de la Santé (IRDES), Paris, Frane; 2 Institut National d Etudes Démographiques (INED), Paris, Frane Correspondene to: Dr Florene Jusot, IRDES, 10 rue Vauvenargues Paris, 75018, Frane; jusot@irdes.fr Aepted 4 June 2007 This paper is freely available online under the BMJ Journals unloked sheme, see jeh.bmj.om/info/unloked.dtl Job loss from poor health, smoking and obesity: a national prospetive survey in Frane F Jusot, 1,2 M Khlat, 2 T Rohereau, 1 C Sermet 1,2 ABSTRACT Bakground and objetives: Health seletion into unemployment may be either diret or operate by referene to health-related behaviours rather than health per se (indiret seletion). Panel data are desirable to investigate seletion effets, and the two types of seletion proesses may be onurrent. We examine jointly the roles of health and health-related behaviours as preursors of unemployment, in order to disentangle diret from indiret seletion proesses. Design: The data of a multi-round nationally representative health survey in Frane were analysed longitudinally, based on three data olletion rounds: , and Following employees salaried in the private setor and aged years at baseline, we explored through logisti regression the influene of nonoptimal self-rated health, smoking and obesity on the risk of being found unemployed 4 years later. Results: After adjustment for self-rated health, obesity was found to be a signifiant preursor of unemployment in women, and heavy smoking had that role in men. After adjustment for smoking and obesity, poor health at baseline was found to be a signifiant preursor of unemployment in both genders. Conlusion: Those findings onfirm the intrinsi role of poor health and of health-related behaviours as preursors of unemployment, with gender-speifi patterns for the latter. Publi poliy presriptions regarding employees protetion from job inseurities should integrate appropriate aommodations of health limitations, and the personal fators underlying unfavourable work and health behaviours should be investigated, in order to thwart indiret seletion phenomena. Unemployment and various types of flexible employment are assoiated with bad health outomes, 1 2 and this is often interpreted within the soial ausation hypothesis. From there, the idea that job inseurities should be a point of onern for publi health poliies is gaining ground in the epidemiologial literature. 3 And yet, in order to elaborate well founded poliy presriptions, the role of the seond faet of the ausation-seletion framework that is, the seletion hypothesis, should be given more onsideration. On one hand, the diret seletion hypothesis assumes that health exerts a diret influene on employment, with people oming to be unemployed or remaining unemployed beause of a disease. 4 Indeed, people who develop hroni illnesses during their working life may beome less able to ope with the demands of the job, and this may be ompounded by the time-onsuming use of medial servies and the frequent sik leaves. They are therefore more likely to lose their job or leave it, in relation to their health limitations. Even if they lose their job for other reasons, they may have more diffiulties than the others in regaining employment. 5 On the other hand, the idea that seletion might operate by referene to health behaviours rather than health itself direts attention to another explanatory mehanism, whih is that of indiret seletion. 6 The indiret seletion proess assumes that ertain personality harateristis may influene both employability through work behaviour and health through health behaviour, therefore generating a fallaious assoiation between unfavourable health behaviours and unemployment. 7 It has also been argued that the true ausal relationship is that by whih the personality trait affets both health and unemployment separately, 7 although those traits are rarely spelt out in exat terms. 4 Therefore, one way to investigate indiret seletion is to onsider that health behaviours an be taken as proxy of the unobservable personal harateristis upon whih they depend. The santy longitudinal studies on health seletion have provided interesting evidene. In The Netherlands, 8 Norway, 9 10 Finland, 11 Sweden, 12 Canada, 13 the United States, 14 Germany 15 and some regions of Frane, 16 mobility out of employment was found to be health seletive. However, most of those studies have investigated transitions out of employment (that is, into either unemployment or inativity), and only a few have speifially targeted the group of people losing their job from poor health, and willing to work again. In a few other prospetive studies, an assoiation has been reported between smoking and subsequent unemployment, and between obesity and unemployment. 21 The interpretation of those findings is not straightforward, as: (1) the health fators underlying the transition into eonomi inativity are likely to be more severe and perhaps different from those attahed to mobility into unemployment, and therefore the piture is likely to be somewhat mixed, and (2) behavioural risk fators and health are likely to be strongly orrelated, and therefore the diret role of health on transitions in and out of employment is diffiult to disentangle from the indiret role of behavioural risk fators, as eah may onfound the other. The aim of this study is to analyse jointly the roles of health, smoking and obesity as preursors of unemployment in Frane, in order to extriate the ontribution of diret health seletion to the assoiation between unemployment and ill health from that of indiret seletion. For this purpose, we build on the National Health, Health Care and J Epidemiol Community Health: first published as /jeh on 13 Marh Downloaded from on 14 November 2018 by guest. Proteted by opyright. 332 J Epidemiol Community Health 2008;62: doi: /jeh

2 Insurane Survey, whih is the only follow-up health survey available at the national level, and on self-rated health as an indiator of global health, and investigate the transitions based on survey data olleted during the 1990s for employees salaried in the private setor and aged years at baseline. METHODS Data olletion The National Health, Health Care and Insurane Survey (ESPS: Enquête sur la Santé et la Protetion Soiale ) is a multi-round health interview survey of all residents in households with at least one member inluded in the EPAS ( éhantillons permanents d assurés soiaux ), whih is an ongoing random sample of benefiiaries of the national health insurane funds. This survey overs more than 95% of Frenh households, and is oordinated by the Institute for Researh and Information in Health Eonomis (IRDES) in Paris. 22 The ESPS was designed as a snapshot survey of the population health and healthare onsumption at regular intervals, and inluded every year one-fourth of the EPAS between 1992 and 1997, and one-half every two years between 1998 and At eah survey round, about 40% of the households in the EPAS files ould not be reahed, owing partly to inevitable losses aused by inomplete data and wrong or outdated addresses. Seventy per ent of the households that were reahed agreed to partiipate in the survey, and the partiipation rate inreased over time. 22 Between data olletion rounds, a majority of the subjets remained in the EPAS sample, while new subjets were inluded, and others were removed. Given this sheme, respondents be followed up every four years, with a maximum of three partiipations between 1992 and On average, the follow-up rate between onseutive survey rounds was lose to 55%, and we found that eduational level had no influene on the likelihood of follow-up; neither had any of the health variables, exept smoking, with heavy smokers being less likely to be followed up. 23 We are fousing on the transitions from employment at time t to unemployment at time t+4 years, in relation with a set of independent variables observed at time t. In the survey questionnaire, the response options for oupational status were ative, unemployed and inative (inluding retired, students, military ontingent, housewives and reipients of disability pensions). People who had delared they were ative had to speify whether they were self-employed, salaried in the private setor or salaried in the publi setor. Self-employed (independent professionals, farm managers, raftsmen and independent shop or business owners) were exluded from the study: indeed, they are neither subjeted to the risk of being fired, nor entitled to unemployment benefits when they ease their professional ativity. In the latter ase, they are expeted to delare that they are inative rather than unemployed. In Frane, employees salaried in the publi setor have permanents jobs and are therefore proteted from unemployment. Below age 30, individuals who lose their job may resume training or formal studies, thereby exiting the labour fore. 24 At the opposite age sale, starting from age 55, individuals who lose their job may be exempted from job seeking while retaining unemployment benefits until they reah pension age and laim their pension entitlement. 25 Therefore, we have seleted the individuals meeting the following riteria: (1) they had to be salaried in the private setor, aged years and with either an indefinite-term or a fixed-term ontrat at the starting point, (2) to be either still salaried or unemployed 4 years later, and (3) to have no missing value for health and behavioural risk fators at baseline. Salaried employees who met the first ondition but were found to be inative or retired at the seond point were exluded. Individuals may meet the inlusion riteria either for only one transition between two onseutive periods, or for both. In the latter ase, one transition out of the two was randomly seleted in order to avoid both the issue of autoorrelation of residuals and the bias in favour of the individuals who had remained employed. The study sample onsisted of 2420 transitions for women and 3287 for men. Health measures Self-rated health was assessed at the first point of eah study period, by asking the respondents to rate their own health on a sale ranging from 0 to 10, and dihotomising the grades into Table 1 Desription of the study sample % No % No Employment status at time t+4 Still employed Unemployed Charateristis at time t Self-rated health Sore >8 (optimal) Sore,8 (non-optimal) Underweight or normal Overweight Obese Smoking Non-smoker Former smoker Smokes up to 1 pak/day Heavy smoker All sample Evidene-based poliy and pratie Salaried employees in the private setor aged years in Enquête sur la Santé et la Protetion Soiale survey, , and J Epidemiol Community Health: first published as /jeh on 13 Marh Downloaded from on 14 November 2018 by guest. Proteted by opyright. J Epidemiol Community Health 2008;62: doi: /jeh

3 Evidene-based poliy and pratie Table 2 Proportion of unemployed at time t+4 years among salaried employees in the private setor aged years at time t in ESPS* survey, , and Charateristis at time t % 95% CI % 95% CI Self-rated health Sore >8 (optimal) to to 5.1 Sore,8 (non-optimal) to to 9.4 Underweight or normal to to 6.3 Overweight to to 6.1 Obese to to 5.6 Smoking habits Non-smoker to to 4.9 Former smoker to to 6.9 Smokes up to 1 pak/day to to 6.4 Heavy smoker to to 11.6 All sample to to 5.7 *ESPS, Enquête sur la Santé et la Protetion Soiale. greater or equal to 8 ( optimal health ) as opposed to lower than 8 ( non-optimal health ). In the 2002 survey round, both this question and the lassi one with five response options (very good, good, average, bad, very bad) were asked, and the orrespondene between the sores of 8 or above and the responses good or very good was quite good. 23 Alohol onsumption was not reorded in the survey. Smoking onsumption was ategorised into non-smoker, smoker of 20 igarettes per day (a pak) or less (light or moderate smoker), smoker of more than 20 igarettes per day (heavy smoker), ex-smoker. Body mass index, alulated from reported height and weight, was ategorised as: less than 25 (underweight or normal), greater or equal to 25 and lower than 30 (overweight), above or equal to 30 (obese). Statistial methods of analysis Logisti regression was firstly used to analyse the relation of behavioural risk fators (smoking, body mass index) to selfrated health at the same time (ross-setional analysis), ontrolling for age (ategorised into 5-year intervals) and eduation level (primary shool; lower seondary; higher seondary; university). The next stage onsisted of an analysis of the influene of behavioural risk fators and self-rated health assessed at time t on employment status at time t+4 years (longitudinal analysis). The ontrol variables in the latter analysis were the following soiodemographi harateristis assessed at time t: age (ategorised into 5-year intervals), type of job ontrat (fixed-term; indefinite-term), eduational level (primary shool; lower seondary; higher seondary; university), type of household omposition (ouple with hildren; lone person; lone mother or father with hildren; ouple without hildren; other), and presene of hildren aged less than 6 years in the household. This approah led to the analysis of (1) the assoiation between self-rated health at time t and employment status at time t+4 years, adjusted for smoking habits, obesity and soiodemographi variables (at time t); (2) the assoiation between smoking habits at time t and employment status at time t+4 years, adjusted by self-rated health, obesity and soiodemographi variables (at time t); and (3) the assoiation between obesity at time t and employment status at time t+4 years, adjusted by self-rated health, smoking habits and soiodemographi variables (at time t). All the analyses were made separately for men and women. RESULTS Self-ratings of health were less favourable among women than among men, with no differene with respet to obesity, and a substantially higher proportion of heavy smokers among men than among women. More women than men were found to be unemployed 4 years after initial observation (table 1). The proportion of unemployed at time t+4 years was higher among women and men reporting non-optimal self-rated health and among obese women (table 2). The first regression analysis foused on the relation of behavioural risk fators (body weight and smoking) to selfrated health at time t (ross-setional analysis). Both attributes turned out to be strongly related to health outomes (table 3). Obese women were muh more likely to report non-optimal health than those in the normal range, and men too, but to a lesser extent. Regular smoking was assoiated with a definite health disadvantage, but in men only, and that disadvantage extended to all types of smokers: former smokers, smokers of up to one pak per day and heavy smokers. The seond regression analysis (table 4) foused on the relation of non-optimal health, regular smoking and obesity of employed individuals at baseline time t to unemployment observed at time t+4 years (longitudinal analysis), based on a regression model inluding both the soiodemographi fators and the health-related fators. After adjustment for self-rated health and smoking habits, obesity was a signifiant preursor of unemployment for women only. who were heavy smokers at time t were muh more likely to have moved into unemployment at time t+4 years than the non-smokers, after adjustment for self-rated health and obesity. Lastly, a nonoptimal self-rated health (graded as less than 8) was assoiated with signifiantly elevated odds ratios in women and in men, after adjustment for smoking habits and obesity. DISCUSSION In Frane, the literature on unemployment and health is very limited, and the most striking feature to date is the onsiderable mortality disadvantage of the unemployed, in omparison with other ountries. In order to shed some light on the fators underlying this disadvantage, this study foused on health-related attributes and poor health, and their speifi role as preursors of mobility from employment to unemployment four years later. Obese men and women and male heavy smokers were found to be muh more likely to J Epidemiol Community Health: first published as /jeh on 13 Marh Downloaded from on 14 November 2018 by guest. Proteted by opyright. 334 J Epidemiol Community Health 2008;62: doi: /jeh

4 Table 3 Odds ratios assoiated with non-optimal self-rated health for behavioural risk fators at baseline time t (ross-setional analysis) Independent variables Underweight or normal Overweight 1.2 (0.9 to 1.5) 1.0 (0.8 to 1.2) Obese 2.7 (1.9 to 3.8) 1.6 (1.2 to 2.2) Smoking habits Never-smoker Former smoker 1.0 (0.8 to 1.3) 1.3 (1.0 to 1.6) Smokes up to 1 pak/day 0.9 (0.7 to 1.2) 1.4 (1.1 to 1.8) Heavy smoker 1.2 (0.6 to 2.5) 2.2 (1.6 to 3.1) Private setor employees aged years in Enquête sur la Santé et la Protetion Soiale survey, , and *Adjusted for age and eduational level. report non-optimal health in the ross-setional analyses. Analysing the data longitudinally, we find that non-optimal health, obesity in women and heavy smoking in men are independently assoiated with subsequent unemployment. One asset of the urrent study is its prospetive nature: the health and behavioural risk fators of the individuals were olleted while they were employed, and related to their employment status four years later. Some of the study limitations may have influened our findings. One issue is the lak of information on the events that have ourred during the 4-year time period between the two survey rounds. This type of inompleteness of the biographial information is expeted to onfuse the piture and therefore to lead to an underestimation of the assoiations between labour market trajetories and health. The same reasoning applies to the lak of information on the irumstanes of job loss (resignation, dismissal or layoff). Given the design of ESPS as a snapshot survey of the population health and healthare onsumption, no speial efforts were made to trak the respondents from one round to the other, and this explains the relatively low follow-up rate (about 55%). One onern is the potential attrition bias, even though our investigation of the underlying fators did not support the existene of differential attrition aording to health. In fat, there were many reasons behind losses to followup (unavailability or absene of respondent at time of visit, respondent move, entire household move, outdated information in the administrative files, refusals), and those reasons were not neessarily related to health. However, heavy smokers were found to be more likely to be lost, and so were unemployed individuals. We may therefore antiipate an under-representation of heavy smokers having lost their job in the sample, leading inevitably to a weakening of the strength of the assoiation between smoking and subsequent unemployment. Also, the lak of information on alohol onsumption is regrettable, as the level of onsumption by men is partiularly high in Frane in omparison with other European ountries. 30 Given that in Frane alohol onsumption is more frequent among smokers, 31 a onfusion between the role of tobao and that of alohol is not unlikely. As expeted, we did find an assoiation between smoking habits and self-rated health on one hand, and obesity and selfrated health, on the other hand, and this aords with the literature. 32 We next analysed jointly the survey information on self-rated health, smoking habits and obesity in order to gain additional insights into the health-related proesses of job loss. By inluding the three fators in the same analysis, we were able to eliminate possible onfounding, and extriate the speifi influene of eah of them on the transition to unemployment. Traditionally, the role of ill health is onsidered to be a refletion of diret seletion, and smoking or obesity as the refletion of indiret seletion. The mehanisms generally hypothesised for health seletion are a redution of produtivity in relation to either ill health or ertain health behaviours. West has, however, viewed health seletion as a profoundly soial Table 4 Odds ratios assoiated with being unemployed at time t+4 years for self-rated health and behavioural risk fators at baseline time t (longitudinal analysis) Independent variables Evidene-based poliy and pratie Self-reported health Sore > Sore,8 1.7 (1.2 to 2.3) 1.5 (1.1 to 2.2) Underweight or normal Overweight 1.3 (0.9 to 1.9) 0.8 (0.6 to 1.2) Obese 2.0 (1.2 to 3.4) 0.5 (0.2 to 1.0) Smoking habits Non-smoker Former smoker 0.8 (0.5 to 1.2) 1.4 (0.9 to 2.1) Smokes up to 1 pak/day 0.9 (0.6 to 1.4) 1.3 (0.8 to 2.0) Heavy smoker 1.7 (0.6 to 4.8) 1.8 (1.0 to 3.3) Private setor employees aged years in Enquête sur la Santé et la Protetion Soiale survey, , and *Adjusted for age, eduational level, type of job ontrat, household omposition and presene of hildren of less than 6 years in the household. J Epidemiol Community Health: first published as /jeh on 13 Marh Downloaded from on 14 November 2018 by guest. Proteted by opyright. J Epidemiol Community Health 2008;62: doi: /jeh

5 Evidene-based poliy and pratie What is already known on this subjet Some studies have demonstrated that employees reporting poor health are more at risk of job loss, and this is interpreted as diret health seletion. Similarly, smokers and obese workers were found to fae the same type of risk, and this is interpreted as indiret health seletion. Possible onfounding between the effets of health and those of health behaviours has not been properly onsidered in those studies. proess of disrimination by health. 6 In fat, the word disrimination omes from the Latin disriminare, whih means to distinguish between, and disrimination is oneptualised as an ation based on prejudie resulting in unfair treatment of people. Distintions between people that are based just on individual merit (suh as personal ahievement, skill or ability) are generally not onsidered soially disriminatory. Disrimination involves treating someone less favourably beause of the possession of a prohibited (undesirable) attribute (among whih disease or disability), regardless of his/her produtivity. West suggests a number of possible dimensions of disrimination, and among them physial attrativeness. 6 In our soiety, attrativeness for women is losely linked to slimness, and therefore obese women may suffer disrimination in the labour market beause they do not live up to the gendered beauty norms for women in soiety. 19 Disrimination with regard to smoking habits is less likely, although it may be hypothesised in some ontexts. Conerning tobao, we did find that employed men who reported that they smoked heavily were muh more at risk of being found to be unemployed 4 years later. In order to onsider residual onfounding from severe tobao-related diseases whih may have ourred between the survey rounds, we have re-run the analyses after inluding self-rated health at time t+4 years among the ovariables. The findings remained essentially the same, whih provides support for the hypothesis of an intrinsi role of tobao onsumption. A statistial assoiation between smoking and subsequent unemployment has been reported in quite a few prospetive studies In Frane, one study has provided evidene in favour of a negative assoiation between smoking and upward oupational mobility in the Gazel ohort of male employees of the national eletriity and gas ompany. 33 In the United States, Ryan et al 17 observed that igarette smoking at the time of hire was assoiated with elevated rates of aidents, absene, disipline, and saking among postal workers, and hypothesise that smoking ould indiate risk taking and suseptibility to aidents. This is somewhat onsistent with the onlusion by Waldron and Lye, 34 that ertain personal harateristis or early experienes influened both smoking adoption and adult unemployment. Lastly, in a study of youth unemployment in Frane, 24 the author onlude that there is a subpopulation that is exluded from the market of long-term, steady jobs, and he relates this situation to personal harateristis determining some kind of intrinsi employability. Another health-related variable that stands out in our study is obesity, whih is inreasingly reognised as a publi health problem in Europe. 35 The prevalene of this ondition is known to vary greatly aording to soial and eonomi fators, but in gender-speifi ways. In women, studies in the developed world have What this study adds Using prospetive data from Frane, we onduted a joint analysis of the influene of health, smoking habits and obesity of employees at a given time on the probability of being unemployed 4 years later, in order to separate diret from indiret seletion. We found onurrent evidene for diret and indiret seletion proesses: (1) non-optimal self-rated health is assoiated with job loss in men and women; (2) there are gender-speifi assoiations for health-related behaviours, with a signifiant role of obesity in women and of heavy smoking in men. Poliy impliations The findings onerning diret seletion support the implementation of properly targeted measures to aommodate the health limitations of workers and improve their employability. The findings onerning indiret seletion support the development of researh into the personal traits underlying behaviours related to both the health and work spheres, and into disrimination proesses operating in the labour market. onsistently reported a powerful inverse relation between soioeonomi position and employment status and obesity: the higher the soial status, the rarer obesity. For men, the assoiation is less onsistent, as it an be positive, negative or absent We found that obese women faed an odds ratio for mobility out of employment of about 2, whih is onsiderable. though may be related to a previous history of unemployment, as ertain subpopulations experiene longterm exlusion from the labour market. 24 In the literature, longitudinal evidene on obesity and employment is sare. The only piee of evidene available in Frane is based on retrospetive information from the 2003 Deennial Health Survey, in whih obese men and women had a muh higher perentage of working years spent unemployed and a muh lower probability of regaining employment than the non-obese men and women. 21 The authors onlude in favour of a signifiant stigmatisation, and even disrimination against obese people, to be attributed to the relative rarity of this ondition in Frane, espeially in omparison with the United States. In a ohort study from Sweden, being overweight at age 16 and at age 21 was related to future working-lass position among women only, whih is interpreted as gendered disrimination, based on the soietal norms for female bodies in our soieties. 19 Conerning the role of ill health, we found that non-optimal self-rated health was a signifiant preursor of unemployment for men and women, with odds ratios around 1.5, after adjustment for smoking and obesity. Those findings aord with prospetive studies of health and subsequent unemployment or unstable employment in Finland, 11 Canada 13 and Sweden, 12 and we further demonstrate that the role of health is intrinsi and not onfounded by indiret seletion proesses operating through health behaviours. A priori, women may be more vulnerable than men in the presene of ill health, as they have to exerise multiple roles. 38 In Frane partiularly, women have a higher partiipation in the labour market than in other J Epidemiol Community Health: first published as /jeh on 13 Marh Downloaded from on 14 November 2018 by guest. Proteted by opyright. 336 J Epidemiol Community Health 2008;62: doi: /jeh

6 European ountries. 39 We did not find any signifiant differene in the amount of health seletion among women ompared to men, and neither did van de Mheen et al in The Netherlands, 8 onsidering the role of less-than-good self-pereived health in assoiation with subsequent unemployment or inativity. In Germany, Arrow reported that health seletion only onerned women and foreign men, onsidered as vulnerable groups. 15 The findings for men and women may be diffiult to ompare; in our survey, about one-quarter of the women reported non-optimal health, as opposed to one-fifth of the men, and this may reflet different pereption of health and diseases. 40 In addition, women are subjeted in Frane to other fators of disadvantage in the labour market, as they onsistently have higher unemployment rates than men, and are muh more frequently involved in parttime jobs, 39 whih may be viewed as an adaptation to the multiple role onstraints. Publi poliy reommendations onerned with workers protetion from job inseurities should inlude proper onsideration of the disadvantage in the labour market of those with worse health and health risks. The handling of diret seletion issues involves the implementation of properly targeted measures to aommodate the health limitations of workers and improve their employability. As far as indiret seletion is onerned, more researh is urgently needed to gain insight into the ommon denominators of health and work behaviours, and in the long run develop appropriate programmes to help individuals towards a better adjustment in the labour market and a safeguarding of their health apital. Disrimination phenomena against obese women or individuals suffering from health problems should also be more thoroughly investigated. Detailed analyses of health trajetories and the development of behavioural risk fators in relation to employment trajetories would provide useful insight into the omplexities of individuals experienes. Aknowledgements: We would like to thank Dr Arhana Singh-Manoux for very helpful omments on a first draft of this manusript. Funding: This study was funded by the Frenh Ministry of Health (DREES), the Frenh Ministry of Labour (DARES), the Researh Fund of La Poste. Competing interests: None. REFERENCES 1. Benah J, Benavides F, Platt S, et al. The health-damaging potential of new types of flexible employment: a hallenge for publi health researhers. Am J Publi Health 2000;90: Kasl S, Jones B. The impat of job loss and retirement on health. In: Berkman L, Kawahi I, eds. Soial epidemiology. Oxford: Oxford University Press, 2000: Bartley M. Job inseurity and its effet on health. J Epidemiol Community Health 2005;59: Bartley M, Ferrie J. Glossary: unemployment, job inseurity, and health. J Epidemiol Community Health 2001;55: Rioux L. Le RMI: entre redistribution et initations. Reherhe d emploi et insertion professionnelle des alloataires du RMI. Eon Stat 2002; : West P. Rethinking the health seletion explanation for health inequalities. So Si Med 1991;32: Bartley M, Ferrie J, Montgomery S. Living in a high-unemployment eonomy: understanding the health onsequenes. In: Marmot M, Wilkinson R, eds. Soial determinants of health. Oxford: Oxford University Press, 2000: Van de Mheen H, Stronks K, Shrijvers CTM, et al. The influene of adult ill health on oupational lass mobility and mobility out of and into employment in The Netherlands. So Si Med 1999;49: Elstad JI, Krokstad S. Soial ausation, health-seletive mobility, and the reprodution of soioeonomi health inequalities over time: panel study of adult men. So Si Med 2003;57: Evidene-based poliy and pratie 10. Mastekaasa A. Unemployment and health: seletion effets. J Community Appl So Psyhol 1996;6: Virtanen P, Vahtera J, Kivimäki M, et al. Labor market trajetories and health: a fouryear follow-up study of initially fixed-term employees. Am J Epidemiol 2005;161: Lindholm C, Burström B, Diderihsen F. Class differenes in the soial onsequenes of illness? J Epidemiol Community Health 2002;56: Magee W. Effets of illness and disability on job separation. So Si Med 2004;58: MDonough P, Amik BC 3rd. The soial ontext of health seletion: a longitudinal study of health and employment. So Si Med 2001;53: Arrow J. Estimating the influene of health as a risk fator on unemployment: a survival analysis of employment durations for workers surveyed in the German soioeonomi panel ( ). So Si Med 1996;42: Saurel-Cubizolles M-J, Bardot F, Berneron B, et al. Etat de santé perçu et perte d emploi. (Self-pereived health and job loss). In: Cassou B, Buisset C, Brugere D, et al., eds. Travail, santé et vieillissement. Relations et évolutions ave l âge. Toulouse: Otares Editions, 2001: Ryan J, Zwerling C, Jones M. Cigarette smoking at hire as a preditor of employment outome. J Oup Environ Med 1996;38: Leino-Arjas P, Liira J, Mutanen P, et al. Preditors and onsequenes of unemployment among onstrution workers: prospetive ohort study. BMJ 1999;319: Hammarström A, Janlert U. Health seletion in a 14-year follow-up study a question of gendered disrimination. So Si Med 2005;61: Montgomery SM, Cook DG, Bartley MJ, et al. Unemployment, igarette smoking, alohol onsumption and body weight in young British men. Eur J Publi Health 1998;8: Paraponaris A, Saliba B, Ventelou B., weight status and employability: empirial evidene from a Frenh national survey. Eon Hum Biol 2005;3: Allonier C, Dourgnon P, Rohereau T. L Enquete Santé Protetion Soiale 2004, un outil d analyse pluridisiplinaire de l aes aux soins, de la ouverture maladie et de l etat de sante des Français. Questions d Eonomie de la Sante 2006;105: Jusot F, Khlat M, Rohereau M, et al. Les itinéraires professionnels en relation ave la santé: une exploitation longitudinale de l enquęte sur la santé et la protetion soiale de l IRDES. rap1662.pdf(2007). 24. Lollivier S. Réurrene du hômage dans l insertion des jeunes: des trajetoires hétérogenes. Eon Stat 2000;334: Barnay T, Jeger F. Quels dispositifs de essation d ativité pour les personnes en mauvaise santé? Questions d éonomie de la Santé 2006;108: Sermet C, Khlat M. La santé des hômeurs en Frane: revue de la littérature. Rev Epidemiol Sante Publique 2004;52: Mesrine A. Les différenes de mortalité par milieu soial restent fortes. Données Soiales. INSEE 1999: Mesrine A. Overmortality assoiated with unemployment and eonomi inativity in Frane sine European Conferene on Population La Haye. 29. Mesrine A. La surmortalité des hômeurs: un effet atalyseur du hômage? Eon Stat 2000;334: Cavelaars A, Kunst A, Makenbah JP. Soio-eonomi inequalities in risk fators of morbidity and mortality in the European Community: an international omparison. J Health Psyhol 1997;2: Com-Ruelle L, Dourgnon P, Jusot F, et al. Une omparaison de deux enquetes en population générale. Paris: IRDES Report No 1600, Manderbaka K, Lundberg O, Martikainen P. Do risk fators and health behaviours ontribute to self-ratings of health? So Si Med 1999;48: Ribet C, Zins M, Gueguen A, et al. Oupational mobility and risk fators in working men: seletion, ausality or both? Results from the Gazel study. J Epidemiol Community Health 2003;57: Waldron I, Lye D. Employment, unemployment, oupation, and smoking. Am J Prev Med 1989;5: Sarlio-Lahteenkorva S, Lahelma E. The assoiation of body mass index with soial and eonomi disadvantage in women and men. Int J Epidemiol 1999;28: Gutierrez-Fisa JL, Regidor E, Rodriguez C. Eonomi and soial fators assoiated with body mass index and obesity in the Spanish population aged years. Eur J Publi Health 1995;5: Rosmond R, Bjorntorp P. Psyhosoial and soio-eonomi fators in women and their relationship to obesity and regional body fat distribution. Int J Obes Relat Metab Disord 1999;23: Khlat M, Sermet C, Le Pape A. s health in relation with their family and work roles: Frane in the early 1990s. So Si Med 2000;50: Essafi CA, Buffeteau S. L ativité féminine en Frane: quelles évolutions réentes, quelles tendanes pour l avenir? Eon Stat 2006; : Idler E, Benyamini Y. Self-rated health and mortality: a review of twenty-seven ommunities. J Health So Behav 1997;38: J Epidemiol Community Health: first published as /jeh on 13 Marh Downloaded from on 14 November 2018 by guest. Proteted by opyright. J Epidemiol Community Health 2008;62: doi: /jeh