History of labour market attachment as a determinant of health status: a 12-year follow-up of the Northern Swedish Cohort

His o y o labou ma ke a achmen as a de e minan o heal h s a us: a 12-yea ollow-up o he No he n Swedish Coho Anna-Ka in Waene lund, 1 Pe E Gus a sson, 1 Anne Hamma s öm, 1 Pekka Vi anen 2,3 To ci e: Waene lund A-K, Gus a sson PE, Hamma s öm A, e al. His o y o labou ma ke a achmen as a de e minan o heal h s a us: a 12-yea ollow-up o he No he n Swedish Coho . BMJ Open 2014;4:e004053. doi:10.1136/bmjopen-2013- 004053 ▸P epublica ion his o y o his pape is a ailable online. To iew hese iles please isi he jou nal online (h p://dx.doi.o g/10.1136/ bmjopen-2013-004053). Recei ed 18 Sep embe 2013 Re ised 14 Janua y 2014 Accep ed 16 Janua y 2014 1 Depa men o Public Heal h and Clinical Medicine, Family Medicine, Umeå Uni e si y, Umeå, Sweden 2 School o Heal h Sciences and Ins i u e o Ad anced Social Resea ch, Uni e si y o Tampe e, Tampe e, Finland 3 Depa men o Public Heal h and Clinical Medicine, Epidemiology and Global Heal h, Umeå Uni e si y, Umeå, Sweden Co espondence o D Anna-Ka in Waene lund; annaka in.wae[email p o ec ed] ABSTRACT Objec i e: The p esen s udy aims a using ajec o y analysis o measu e labou ma ke a achmen (LMA) o e 12 yea s and a examining whe he labou ma ke acks ela e o pe cei ed heal h s a us. Design: Da a we e e ie ed om a 26-yea p ospec i e coho s udy, he No he n Swedish Coho . Se ing and pa icipan s: All nin h g ade s uden s (n=1083) wi hin he municipali y o Luleå in no he n Sweden we e included in he baseline in es iga ion in 1981. The as majo i y (94%) o he o iginal coho pa icipa ed a he ou h ollow-up. In his s udy, 969 pa icipan s we e included. Measu es: Pe cei ed heal h s a us (psychological dis ess and non-op imal sel - a ed heal h) a age 42 and he da a ob ained om ques ionnai es. Resul s: We ha e iden i ied ou acks in ela ion o LMA ac oss he 12-yea pe iod: ‘pe manen ’,‘high le el’,‘s eng hening’and ‘poo le el’o a achmen . LMA his o y ela es o psychological dis ess. High le el (OR 1.55 (95% CI 1.06 o 2.27)), s eng hening (OR 1.95 (95% CI 1.29 o 2.93)) and poo a achmen (OR 3.14 (95% CI 2.10 o 4.70) in ol e highe OR o psychological dis ess compa ed wi h pe manen a achmen . The o e all p alue emained signi ican in he inal model (p=0.001). Analyses ega ding non- op imal sel - a ed heal h displayed a simila pa e n bu his was no signi ican in he inal model. Conclusions: Ou esul s sugges ha heal h s a us in mid-li e, pa icula ly psychological dis ess, is ela ed o pa e ns o LMA his o y, o a la ge pa independen ly o o he social isk ac o s and p e ious heal h. Conside a ion o he e ogenei y and ime in LMA migh be impo an when analysing associa ions wi h pe cei ed heal h. INTRODUCTION Du ing he pas h ee decades o so, demands o a mo e flexible wo k o ce ha e inc eased all o e he wo ld. Flexibili y in he labou ma ke is eflec ed, o example, in he p opo ion o empo a y employees, which in ecen yea s has a ied be ween 13% and 15% o he o al wo king popula- ion in Eu ope. 1 To he indi idual, he conse- quence o flexible employmen is ha a achmen o a wo kplace is weak and com- monly in e up ed by unemploymen and o he episodes ou o wo k. In o de o cap u e he o al spec um o employmen ela ions, wi h ega d o quali y and quan i y, we ha e chosen o wo k wi h he concep o ‘labou ma ke a achmen ’(LMA) in his s udy. In ou defini ion o he LMA spec- um, we ha e included pe manen , non- pe manen employmen , unemploymen and hose who a e exemp ed om wo king. 2–4 Some ela ions be ween labou ma ke s a us and heal h s a us a e widely accep ed: pe manen employmen is conside ed bene- ficial o heal h, while unemploymen is known o ha e ad e se heal h e ec s 5–7 and being ou o he labou ma ke in he long e m is commonly due o poo heal h. 8 Bu when i comes o he heal h implica ions o labou ma ke s a us, be ween he ex emes o he LMA spec um, he e idence is much sca ce . P e ious s udies ha e p oposed ha empo a y employmen could be a isk o poo psychosocial wo k cha ac e is ics, 9 wi h empo a y employees mo e commonly expe iencing job insecu i y and ha ing a low S eng hs and limi a ions o his s udy ▪Using longi udinal da a wi h excep ionally high esponse a e. ▪Applying ajec o y analysis which only has p e- iously been spa sely used in his ield. ▪P o ides a uniquely ich and de ailed da a on empo a y con ac s o e a long pe iod o ime. ▪Relies on sel - epo ed da a. ▪Used a ela i ely small sample. Waene lund A-K, Gus a sson PE, Hamma s öm A, e al.BMJ Open 2014;4:e004053. doi:10.1136/bmjopen-2013-004053 1 Open Access Resea ch g oup.bmj.com on June 30, 2016 - Published by h p://bmjopen.bmj.com/Downloaded om cash ma gin, 10 which a e some o he ac o s ha can be po en ial pa hways linking weak LMA o poo heal h. We ha e, o example, shown ha long- e m empo a y employees o en expe ience di ficul ies wo king ull- ime (unde employmen ) which influences hei financial si ua ion in a nega i e way 11 and can hus be a sou ce o wo y. 12 Poo heal h can also be media ed h ough job s ain, al hough i seems as i his g oup is a ec ed mo e by limi ed influence o con ol a he han high demands. 13 Mode a e LMA has been highligh ed as po en ially ha m ul o heal h s a us, bu esea ch on he opic is di e gen . 14 While some s udies epo wo se sel - a ed heal h among empo a y employees, 15 a ew s udies e en sugges be e sel - a ed heal h among fixed- e m employees. 16 Some s udies also indica e ha de e io a ion in sel - a ed heal h may no be obse able un il he a achmen is qui e weak. 17 A majo answe o he mixed esea ch esul s may lie in he inhe en di fi- cul y in cap u ing he changeable na u e o employ- men . Measu emen o he employmen si ua ion a one o a ew poin s in ime, as in mos o he p e ious s udies, 18 19 may o e look he ac ual exposu es o di e - en posi ions. Heal h e ec s o empo a y employmen could su ace a e se e al yea s o accumula ion, as we ha e shown in a p e ious epo , 20 o e en when se e al yea s ha e passed a e he exposu e. 21 Mo eo e , i is possible ha i is he chain o passages in di e en employmen s, he ajec o y o LMA, ha con ibu es o heal h s a us; o example, becoming mo e s ongly a ached o e ime could con ibu e o heal h s a us di - e en ly han becoming less a ached o he labou ma ke e en i he bo h g oups hold same o al exposu e. The e a e a lo o s udies abou pa icula labou ma ke s a us as p edic o s o heal h. An inhe en p oblem o such s udies is he amoun o exposu e: c oss-sec ional in o ma ion abou s a us includes g ea a ia ion, a a iable based on he du a ion o he ongoing s a us is bound o use c oss-sec ional heal h da a, and in p ospec i e ollow-up se ings he e may be pe iods in se e al s a us du ing he ollow-up. The p oblem is o special impo ance in esea ch on he heal h e ec s o a ypical employmen , as he e a e a wide ange o s a us be ween pe manen employmen and o e unemploymen . To add ess his opic, we ha e in oduced a sco e ha sums up he exposu e o di e - en ypes o non-pe manen employmen du ing he ollow-up. 20 The sco e, howe e , does no ake in o accoun he iming o he exposu es. One way o cap u e he s a us chains, o he passages in and he ansi ions be ween di e en labou ma ke posi ions, is p o ided by ajec o y analysis. Applying his me hod wi h a ou - class esponse a iable (pe manen employmen , non- pe manen employmen , unemploymen and ou o he labou o ce), he membe s o he No he n Swedish Coho ha e been clus e ed in o six di e en ‘LMA acks’. 2 Fo he ajec o y analysis o he p esen s udy, LMA was measu ed by a 10-class indica o , in o de o a icula e in mo e de ail he ajec o ies o non- pe manen employmen and hei associa ion wi h heal h. The ew a ailable s udies abou labou ma ke ajec- o ies in he field o public heal h ha e measu ed LMA a a ew ime poin s, o example, be ween one ime poin and ano he 9 and he goal has been desc ip i e a he han analy ic. 19 22 In he p esen s udy, we ha e applied a efined scale wi h ega d o LMA. In he measu e, we ha e included a spec um o a ious ypes o employmen si ua ions and also co e ed a ime-span o 12 yea s. The me hod, ajec o y analysis, 23 o e s a no el way o iden i y di e en ial acks o LMA his o y in ou popula ion coho . A e ob aining a ele an se o acks, ou aim is o examine in which ways hey ela e o non-op imal sel - a ed heal h o pe cei ed psychological dis ess. METHOD Popula ion and p ocedu es The s udy was ini ia ed in 1981 and included all pupils in hei las yea o compulso y school (n=1083), in he medium-sized indus ial own o Luleå in no he n Sweden (95% o he coho we e bo n in 1965). Since 1981 ou addi ional ollow-ups ha e been conduc ed. 24 The a i ion a e in he mos ecen ollow-up om 2007 was low; 94% (n=1005) o he o iginal coho who we e s ill ali e (n=1071) pa icipa ed. Comp ehensi e ques ionnai es, comple ed by he pa icipan s in 1981, 1983, 1986, 1995 and 2007, we e used as he main assess- men me hod. This pape is based on da a om he 1995 and 2007 ollow-ups. P ocedu es o da a collec ion a e desc ibed mo e ex ensi ely elsewhe e. 24 The s udy was app o ed by he Regional E hical Re iew Boa d in Umeå, Sweden. Measu emen s LMA his o y We name he esponse a iable o he ajec o y analysis as LMA. I aims o se e as a concep ual and empi ical ool o so he employmen s a us o he pos indus ial labou ma ke along a con inuum. 25 26 C udely, i ‘ e e s o whe he o no people ha e con inuous employmen (eg, all yea o only pa o i ) and whe he o no hey ha e pe iods o unemploymen ’ 27 A headline le el 28 he e a e ou majo classes o LMA: non-employmen , unemploymen , empo a y employmen and pe manen employmen . Wi hin each class, se e al subclasses can be disce ned. In he p esen s udy, he in e es is ocused on di e en ial empo a y employmen . Seen h ough LMA, empo a y employmen co e s a se o posi ions defined by o mal job con ac s, ega dless o he psy- chological con ac , 29 job commi men 30 o pe cei ed job insecu i y. 31 Pa icipan s’labou ma ke posi ion om 1996 o 2007 was measu ed wi h a ma ix consis ing o columns ep e- sen ing hal -yea pe iods and ows ep esen ing di e en 2Waene lund A-K, Gus a sson PE, Hamma s öm A, e al.BMJ Open 2014;4:e004053. doi:10.1136/bmjopen-2013-004053 Open Access g oup.bmj.com on June 30, 2016 - Published by h p://bmjopen.bmj.com/Downloaded om labou ma ke posi ions. Wi h he ins uc ion ‘Du ing which pe iods ha e you been employed pe manen ly o ha e had some ype o empo a y job con ac o ha e been ou o wo k?’ he esponden s we e p omp ed o choose among 11 esponse op ions o each 6-mon h pe iod: ‘pe manen ly employed’(coded as 10), ‘en ep e- neu ’(9), ‘employed in p ojec ’(8), ‘subs i u e’(7), ‘p o- ba iona y employmen ’(6), ‘on-demand wo ke ’(5), ‘seasonal wo ke ’(4), ‘ empo a y employee o o he easons’(3), ‘in employmen policy measu e’(2), ‘unemployed’(2) and ‘ou o he labou ma ke ’(1). Unemploymen and pa icipa ion in policy measu e we e me ged, and each op ion was coded o a a iable ha exp essed he s eng h o LMA on a scale om 10 o 1. The anking o con ac s has been es ed in p e ious esea ch on he accumula ion o empo a y employmen 20 and was based on A onsson’sco e–pe iphe y model, 32 which p oposes ha he e is a heal h g adien in ela ion o he ype o employmen con ac . The model anks con- ac s based on du a ion, possibili y o on- he-job aining, au onomy and job secu i y. 32 Howe e , we ha e ex ended he anking by also including h ee labou ma ke posi- ions o non-employmen . In he e en o ‘ou o labou ma ke ’, a con inuum o fi e o mo e pe iods (ie, 2.5 yea s) was equi ed: i he e we e ewe han fi e, all pe iods we e coded acco ding o he las labou ma ke posi ion, o pu posely exclude hose on pa en al lea e o s udying om his ca ego y. As he las ca ego y, ‘ou o labou ma ke ’was gi en he lowes sco e in ela ion o LMA. The in en ion o his g oup was ha i was supposed o be con ained by hose mainly on sickness benefi (sup- po ed by analysis, da a no shown), no hose empo a ily no wo king due o pa en al lea e o educa ion. These ope a ions hus yielded a sco e o he s eng h o LMA o each o he 24 successi e hal -yea pe iods. The sco es se ed as he da a o he ajec o y analysis. Indica o s o heal h s a us Psychological dis ess a ages 30 and 42 was measu ed wi h a ques ion ha inqui ed whe he he esponden had expe i- enced in he p e ious yea any o he ollowing symp oms: es lessness, concen a ion p oblems, being wo ied o anxious, palpi a ions, anxie y o panic o o he ne ous p o- blems. Repo ing one o mo e o he six symp oms was coded as 1 and none as 0, equalling a dicho omisa ion in o he qua ile wi h he mos symp oms e sus he es . The ques ion was de i ed om he ‘Su ey o li ing condi ions’. 33 Non-op imal sel - a ed heal h a ages 30 and 42 was mea- su ed wi h one ques ion, ‘How do you a e you gene al heal h?’, wi h esponse op ions: good, a e age o bad. 33 The esponses we e dicho omised in o he qua ile wi h he wo s heal h (a e age o bad) coded as 1, and he es (good) coded as 0. 10 34 Co a ia es Socioeconomic posi ion is a s ong p edic o o heal h. 35–38 LMA has been shown o be ela ed o occu- pa ional class. 39 I is impo an o conside socioeconomic posi ion as a possible con ounde when s udying he ela ionship be ween empo a y employ- men and illness. In his s udy, socioeconomic posi ion (SEP) in 2007 was based on one ques ion abou occupa- ion, which was classified acco ding o he Swedish socio- economic classifica ion o occupa ional ca ego ies. 40 Uppe whi e-colla and sel -employed we e coded as 0, lowe whi e-colla wo ke s we e coded as 1 and blue- colla wo ke s we e coded as 2. Pa ne ship and pa en hood a e also impo an ac o s in ela ion o LMA as hese wo may be pos - poned due o insecu e wo king a angemen , 39 41 which could influence social aspec s ela ed o illness. 39 Ma iage o ha ing a pa ne is also impo an o con- side when s udying LMA, as i can be beneficial o heal h and financial esou ces. 42 Pa en al s a us in 2007 was coded as 0 o hose ha ing child en and as 1 o he childless. Ma i al s a us in 2007 was measu ed wi h one ques ion: ‘A e you ma ied o co-habi ing?’yes was coded as 0 and no as 1. Mo e women han men ha e poo LMA in e ms o empo a y employmen . 43 Gende has been conside ed as an impo an ac o in ela ion o poo LMA and illness, whe e women’s heal h migh be a g ea e isk. 44 Women we e coded as 0 and men as 1. S a is ics The pa icipan s we e clus e ed acco ding o he de elop- men o hei LMA o e he 12 yea s applying ajec o y analysis. 23 The me hod has been es ablished as a way o s udying indi idual de elopmen al cou ses o e age o ime, and o iden i ying dis inc i e g oups o indi idual ajec o ies wi hin he popula ion ha eme ge, ins ead o p edefined c i e ia, om he da a i sel . T ajec o y analysis consis s o h ee s eps. Fi s , he app op ia e p obabilis ic model is chosen o he esponse a iable. The second s ep is o define he deg ee o he polynomial o m o he ajec o ies. Finally, he numbe o he clus e s is decided, employing he s a is ical in o ma ion c i e ia and he ‘common sense c i e ia’wi h espec o he subs ance and aims o he s udy. As a esul , he de elopmen al ajec o - ies wi hin clus e s a e as simila as possible, and ajec o ies be ween clus e s a e as di e en as possible. 45 A indi idual le el, clus e membe ship is dic a ed by he highes calcu- la ed pos e io p obabili y o belonging o a pa icula clus e . T ajec o y analyses we e conduc ed wi h he Mplus p og am package. The analysis allowed o 12 ime poin s, and he fi s and he second hal o e e y second yea we e chosen in o de o ake in o accoun he pos- sible sys ema ic seasonal a ia ion o LMA. We used logis ic eg ession (OR and 95% CI) o es whe he LMA his o y was associa ed wi h non-op imal sel - a ed heal h and psychological dis ess a age 42. Adjus men s we e made o he heal h indica o a age 30, gende , socioeconomic posi ion, pa en al s a us and ma i al s a us. SPSS V.17 was used o he eg ession Waene lund A-K, Gus a sson PE, Hamma s öm A, e al.BMJ Open 2014;4:e004053. doi:10.1136/bmjopen-2013-004053 3 Open Access g oup.bmj.com on June 30, 2016 - Published by h p://bmjopen.bmj.com/Downloaded om analyses. Women and men we e analysed oge he o p ese e powe . RESULTS The da a consis o a 10-class o dinal indica o o LMA, and he o m o p obabili y dis ibu ion o longi udinal sequence o measu emen o such a iable is he mul i- nomial dis ibu ion. The adjus ed Bayesian in o ma ion c i e ion (BIC) alue dec eased om 17435.474 o he linea model o 16656.622 o he quad a ic polynomial model, indica - ing ha he la e can be p e e ed o he o me . O he in o ma ion c i e ia p o ided by he Mplus p og am like- wise poin ed in he same di ec ion. The op imal numbe o ajec o ies was sea ched by checking he solu ions up o 1. The Adj. BIC was used o choosing he solu ion. 46 The Adj. BIC alue (as well as he o he in o ma ion c i e ia) dec eased when he numbe o ajec o ies was inc eased, apidly a he beginning and slowe a he end. F om se en- ajec o y o eigh - ajec o y solu ion, he figu e dec eased om 17856.940 o 17606.402. The ea e he dec ease slowed down. We decided o con inue wi h eigh ajec o ies, as his solu ion p o ided, in addi ion o de ailed depic ion o di e en ial LMA, he oppo uni y o exclude he clus e wi h 0 LMA ha was ou side ou esea ch in e es . Figu e 1 illus a es he ‘labou ma ke acks’based on means o he LMA sco es a each ime poin o he indi- iduals classified in o each ajec o y. Indi iduals on ack 1 (3% o he coho ) also we e excluded, as hey we e mainly disabili y pension ecipien s and hei heal h was poo by defini ion. The ack (6) o ‘pe man- en ’employmen h oughou he ollow-up included mo e han hal o he coho , whe eas he emaining six clus e s we e ela i ely small and he e we e ela i ely simila acks. We collapsed hese six clus e s in o h ee as ollows. A conside able pa (classes 2 and 3) main- ained a con inuously ‘high le el’o a achmen , and abou 12% (classes 5 and 7) displayed a ‘s eng hening’ o a achmen owa ds he end o he ollow-up. The a achmen was pe manen ly weak in abou 1 o 10 (class 8), and a small clus e wi h a U-shaped pa e n (class 4) was also seen; we decided o collapse hese clus- e s and defined hei a achmen as ‘poo ’. In addi ion o being subs an ially g ounded, his collapsing p o ided s a is ical powe o subsequen analyses. Thus, we a i ed a a ou class ‘LMA his o y’ a iable ha com- p ised ‘pe manen ’(class 6), ‘high le el’(classes 2 and 3), ‘s eng hening’(classes 5 and 7) and ‘poo ’(classes 4 and 8) LMA (figu e 2). Table 1 shows he dis ibu ion o he key a iables by LMA his o y. In gene al, LMA his o y wi h less s ong a achmen en ailed poo e heal h a age 42. Conce ning co a ia es, simila pa e ns we e seen o psychological dis ess and non-op imal sel - a ed heal h a age 30, hus indica ing ha heal h s a us di e ed be ween LMA his o y al eady a baseline. The e we e also significan di e ences wi h ega d o socioeconomic posi ion; he p opo ion o blue-colla wo ke s was highes in he g oup wi h poo a achmen o he labou ma ke . The poo a achmen g oup was also mos likely o be li ing wi hou a pa ne (p=0.010). Pa en al s a us did no di e be ween he g oups. The p opo ion o women was highe in s eng hening and poo -le el a achmen , whe eas he pe manen and high-le el a achmen we e domina ed by men (p=0.001). To analyse whe he LMA his o y was ela ed o poo heal h, sel - a ed heal h and psychological dis ess we e eg essed on LMA his o y wi h pe manen a achmen as he e e ence ca ego y, in mul iple logis ic eg ession models ( able 2). The LMA his o y had a significan o e all p alue on bo h heal h ou comes. In unadjus ed analyses wi h psychological dis ess, lowe s eng h o a achmen in ol ed highe odds han he pe manen one ( able 2, model 0; LMA his o y p<0.001). In model 0, we ound inc eased OR in ela ion o lowe a ach- men o he labou ma ke , anging om high le el OR 1.55 (95% CI 1.06 o 2.27), s eng hening OR 1.95 (95% CI 1.29 o 2.93) and poo OR 3.14 (95% CI 2.10 o 4.70). Adjus ing o co a ia es (sociodemog aphic and psychological dis ess a age 30, models 1 and 2, espec - i ely) esul ed in a enua ion o he ORs o hose wi h ‘s eng hening’and ‘poo ’a achmen . Finally, in he ully adjus ed models (model 3), only he high le el and poo a achmen we e significan ly ela ed o psycho- logical dis ess, wi h he ‘poo ly’a ached displaying he nume ically s onges associa ion (OR=2.52 CI 1.59 o 3.98). Al hough bo de line significan , he OR o he ‘s eng hening’(OR=1.57 (95% CI 0.99 o 2.49)) a ach- men was o compa able s eng h o he ‘high le el’ a achmen (OR=1.54 (95% CI 1.01 o 2.35)). While wi h a sligh dec ease in OR along wi h adjus men s o sociodemog aphic ac o s and p e ious heal h, he same pa e n is s ill e iden in he final model, and he o e all p alue emained highly significan (p=0.001). Resul s o non-op imal sel - a ed heal h we e sligh ly less p ominen han o psychological dis ess bu poin ed in he same di ec ion. Unadjus ed models ( able 2, model 0) showed a significan o e all ela ion- ship be ween LMA his o y and sel - a ed heal h (p<0.001), and he LMA his o y wi h ‘s eng hening’ (OR=1.57 95% CI 1.05 o 2.36) and ‘poo ’(OR=2.26 CI 1.53 o 3.34) a achmen had highe odds o ha ing non- op imal sel - a ed heal h. Fo hose wi h ‘high le el’ (OR=1.38 CI 0.96 o 2.00) o a achmen , he isk was nume ically highe , al hough no significan ly, han o hose wi h ‘pe manen ’a achmen . A e adjus men o heal h s a us a age 30 (model 2), he esul s o he ‘s eng hening’g oup we e a enua ed below signifi- cance. In he final model, only he OR o he ‘poo ’ a achmen g oup was significan ly high, while he o e all p alue o he associa ion o LMA his o y wi h non-op imal heal h d opped o bo de line significance (p=0.087). 4Waene lund A-K, Gus a sson PE, Hamma s öm A, e al.BMJ Open 2014;4:e004053. doi:10.1136/bmjopen-2013-004053 Open Access g oup.bmj.com on June 30, 2016 - Published by h p://bmjopen.bmj.com/Downloaded om DISCUSSION In his s udy, we ha e iden ified ou acks o LMA his o y o e 12 yea s: pe manen , high le el, s eng hening and poo . Compa ed wi h he coho wi h ‘pe manen ’a ach- men , we ound a highe p obabili y o psychological dis- ess among h ee coho s wi h non-pe manen LMA. The di e ences we e pa ially a ibu ed o p e ious heal h and o some ex en also o sociodemog aphic a iables; ne e - heless, he o e all con ibu ion o LMA his o y emained significan in he ully adjus ed model. The esul s o non- op imal sel - a ed heal h we e simila bu somewha weake . Much o he p e ious esea ch has sugges ed ha em- po a y employmen is a s epping s one owa ds mo e s able employmen . 47 In he p esen s udy, we ound ha , al hough he g ea e pa o he sample al eady had a ai ly s ong a achmen o we e mo ing owa ds s onge a achmen , one ack emained poo ly a ached o he labou ma ke o e he 12 yea s. Thus, while he s epping s one hypo hesis migh be co ec o he majo i y, a subs an ial mino i y appea s o be apped in an employmen si ua ion o pe manen em- po a iness. Fu he , we ound mo e women han men in his leas a ou able si ua ion. This is p obably explained by he widesp ead use o ‘on-demand employmen con- ac s’in women-domina ed sec o s o business 43 such as ca e and wel a e and educa ion 48 which a e wo indus- ies which oge he s and o 33% o all empo a y con- ac s in Sweden. 11 These esul s can be in e p e ed as showing ha , e en hough Sweden is conside ed one o he op- anked gende -equal coun ies wi h ega d o high labou ma ke pa icipa ion among women, he in e nal labou ma ke is s ill gende -seg ega ed. Ou findings impo an ly sugges ha , in addi ion o he p esen LMA, he LMA his o y is impo an o heal h. In line wi h p e ious esea ch, we ound ha he g oup wi h leas a achmen had he wo s heal h s a us. 15 17 As in ou p e ious epo showing a cumula- i e e ec o empo a y employmen on heal h s a us, 20 he p esen epo u he co obo a es ha Figu e 1 The numbe o coho membe s clus e ed by ajec o y analysis in o eigh acks (cu es based on means o he a achmen a he ime poin s). Figu e 2 Collapsed acks o labou ma ke a achmen his o y; pe manen , high-le el, s eng hening and poo a achmen . Waene lund A-K, Gus a sson PE, Hamma s öm A, e al.BMJ Open 2014;4:e004053. doi:10.1136/bmjopen-2013-004053 5 Open Access g oup.bmj.com on June 30, 2016 - Published by h p://bmjopen.bmj.com/Downloaded om employmen his o y impac s on heal h s a us. Howe e , a mo e nuanced pic u e eme ged when he sepa a e labou ma ke acks we e conside ed. Fi s , e en hose wi h ‘high-le el’a achmen expe i- enced psychological dis ess mo e commonly han hose wi h ‘pe manen ’a achmen . Thus, i appea s ha e en ma ginal exposu e o wha could be labelled ‘LMA s ess’can con ibu e o psychological dis ess o e ime. Second, he g oup wi h ‘s eng hening’a achmen s ill expe ienced poo heal h mo e commonly han hose wi h pe manen a achmen , e en a e adjus - men s o sociodemog aphic a iables. This esul sug- ges s ha ea lie subop imal a achmen has a po en ially ‘sca ing’e ec . I is impo an o poin ou ha in c oss-sec ional o sho - e m longi udinal designs, he ‘s eng hening’g oup would ha e been ega ded as ha ing a ai ly a ou able LMA. This pa icula finding, he e o e, s esses he impo ance o aking he specific pa e ns o he long- e m labou ma ke his o y in o accoun when examining associa ions wi h heal h s a us. Thi d, among hose poo ly a ached o he labou ma ke is whe e we ound he highes ORs o poo heal h. The poo ly a ached wo ke s we e also mo e likely o ha e o he isk ac o s o psychological dis ess, which seemed o pa ly accoun o he associa ion. Fo example, being single in ol es a lowe a e age house- hold income and dec eases he le el o con ol, 49 and poo p e ious heal h could also con ibu e o di ficul ies in ge ing a s able a achmen o he labou ma ke , as hose wi h heal h oubles could be less employable, pa - icula ly o blue-colla occupa ions. 50 The poo ly a ached wo ke s a e also hose wi h mos expe ience o unemploymen , which p e iously has been connec ed o mo ali y and mo bidi y. 551 Thus, he poo men al heal h in his g oup seems o be a esul no only o hei endu ingly un a ou able LMA his o y, bu also o a ange o coexis ing bu densome li e ci cums ances. As desc ibed in he in oduc ion, job insecu i y, financial s ess, lack o ecip oci y, unce ain y and lack o au on- omy a e some o he b oad ange o po en ial pa hways linking empo a y employmen o poo heal h in p e i- ous s udies. 9101552–54 Al hough media ing mechanisms has no been wi hin he scope o his a icle, ou esul s can be unde s ood in he ligh o po en ial pa hways p e- sen ed in p e ious esea ch. Me hodological conside a ions Ou s udy elies on ajec o y analysis, a me hod ha became a ailable in he beginning o he 2000s, la gely by i ue o inc eased compu ing powe . Ins ead o p e- fixing al e na i e de elopmen s, he analysis is seeking de elopmen al ajec o ies ha eme ge om he da a i sel . 17 I gi es each indi idual pos e io p obabili y alues o each ajec o y, and alloca es hem acco ding o he highes p obabili y in o g oups o ca ego ies so ha indi iduals wi hin a g oup a e mo e simila han indi iduals be ween g oups. 55 To ou knowledge, ajec- o y analysis has been used only spa sely in p e ious s udies abou LMA. 56 Ou p e ious esea ch using a co e–pe iphe y s uc u e has only yielded a measu e summa ising he o al accumula ion o pe iphe al employmen o e 12 yea s, 20 whe eas his s udy has added new knowledge o he field by showing when exposu e akes place and has p o ided pa e ns o labou ma ke his o y, and his specific knowledge has hin ed abou po en ial sca ing e ec s o poo LMA (seen in he s eng hening g oup), knowledge ha has p e iously been hidden due o me hodological limi a- ions. In gene al, his s udy ep esen s no el me hodo- logical ideas in a esea ch a ea ha has been obscu ed Table 1 Dis ibu ion (%) o dependen and independen a iables in ela ion o labou ma ke a achmen (LMA) his o y Pe manen , n=550 High le el, n=163 S eng hening, n=126 Poo , n=130 p Value (χ 2 es )T ack 0 1 2 3 Dependen a iables a age 42 Psychological dis ess 26.3 35.7 41.0 52.8 <0.001 Non-op imal sel - a ed heal h 29.1 36.2 39.2 48.1 <0.001 Independen a iables Heal h s a us a age 30 Psychological dis ess 19.4 27.3 36.3 31.9 <0.001 Non-op imal sel - a ed heal h 16.8 29.4 26.6 33.3 <0.001 Sociodemog aphics <0.001 Blue-colla wo ke 33.5 25.8 35.7 47.7 Lowe whi e-colla wo ke 15.7 8.0 18.3 6.2 Uppe whi e-colla wo ke 50.8 66.3 46.0 46.2 Single ma i al s a us 19.1 20.9 22.2 32.6 0.010 No child as pa en al s a us 17.3 16.6 15.1 19.4 0.831 Women 44.5 42.3 61.9 55.4 0.001 All alues a e displayed in pe cen , excep p alues. 6Waene lund A-K, Gus a sson PE, Hamma s öm A, e al.BMJ Open 2014;4:e004053. doi:10.1136/bmjopen-2013-004053 Open Access g oup.bmj.com on June 30, 2016 - Published by h p://bmjopen.bmj.com/Downloaded om by he p ac ice o measu ing he labou ma ke posi ion c oss-sec ionally o wi h sho - e m ansi ions be ween labou ma ke posi ions. The longi udinal design o his s udy has been an asse , as i made i possible o adjus o p e ious heal h, and he eby educed he possibili y o e e se causa ion, and especially as he a i ion a e was kep low. Conce ning he gene alisabili y o his s udy, his coho has been shown by a p e ious assessmen o be ep esen- a i e o Sweden as a whole wi h espec o demog aphic da a. 24 I is plausible ha heal h e ec s o poo LMA ope a e depending on he social con ex . Fo example, heal h implica ions migh be mo e o less e iden depending on s uc u al ac o s such as na ional labou ma ke policies, educa ion sys em and legisla ion. 57 Sweden is pa o he Scandina ian wel a e egimes which a e conside ed o ha e s ong Social Democ a ic alues and go e nmen - unded benefi s du ing episodes o unemploymen . The wel a e s a e Sweden could, he e o e, possibly educe nega i e heal h e ec s o flex- ible employmen , 58 which is in con as o he esul s in ou s udy. Howe e , Swedish unions ha e c i icised he cu en labou ma ke egula ions o being oo libe al ega ding empo a y employmen . Wi h cu en egula- ions, i is possible o hi e a subs i u e o up o 2 yea s and a e ha hi e he same pe son on a gene al em- po a y employmen con ac o up o 2 yea s. This causes a si ua ion whe e people a e a isk o becoming long- e m empo a y. 59 As a esul o his, app oxima ely 10% o all empo a y employees ha e been employed by he same employe o 5 yea s o mo e. The Swedish labou ma ke egula ion could, he e o e, be a eason o he no iceable finding in his s udy, whe e a subs an- ial pa o he wo ke s ollowed emained poo ly a ached o e he 12 yea s which we e examined. Long- e m empo a y employmen could be a u u e p oblem, and also a ele an g oup o s udy u he in u u e esea ch. The sample size was limi ed in his s udy, which is a poin e o po en ial u u e esea ch, which would benefi om analysing da ase s wi h a la ge sample size. Such s udies would be able o s a i y analyses o gende o socioeconomic posi ion, which could en ich he unde s anding o he field o LMA and illness. In his s udy, we ha e ocused on poo LMA as a isk ac o o poo heal h, bu he e a e a ange o o he ci cum- s ances ela ing o p eca ious employmen which could explain he esul s, such as ulne abili y, lack o benefi s, low wages and disempowe men . 44 The e o e, we would ecommend u u e esea ch o elabo a e on o he aspec s o p eca iousness linked o poo LMA, as well as explo ing he alidi y o he phenomena in a di e en con ex , o example, coun ies wi h di e en social and labou policies. This s udy is based on sel - epo ed da a. Equi alen measu es o sel - a ed heal h and psychological dis ess ha e displayed good alidi y by p edic ing u u e mo al- i y and mo bidi y 60 e en a e adjus men s o known heal h isks. 61 Fu he mo e, using e ospec i e measu e- men o LMA his o y could lead o ecall bias. Ne e heless, e ospec i e ques ions abou occupa ional his o y ha e been shown o main ain good quali y in e ms o ag eemen wi h census da a. 62 Fu he , he ac ha we combined acks could be seen as a limi a ion o his s udy. Howe e , in ou unde s and- ing, some o he acks could be combined depending on hei simila mo emen in he labou ma ke . We fi s analysed each o he eigh acks sepa a ely in ela ion o he ou come, o make su e ha none o he combina ions esul ed in blu ing o he esul s (da a no shown). One limi a ion ega ding he me hod was ha , due o ea u es o he p og amme, only hal o he ime poin s (12 o 24 possible ime poin s) could be included in he ajec o y analysis. Howe e , we used he fi s and second hal o e e y second yea o educe he possibili y o sys ema ic e o due o seasonal a ia ions in employmen s a us. Table 2 Logis ic eg ession analyses o wo ou comes a age 42 in ela ion o exposu e o LMA his o y (OR (CI 95%)) T ajec o ies Model 0 Model 1 Model 2 Model 3 Psychological dis ess as ou come Pe manen , n=550 1 1 1 1 High le el, n=163 1.55 (1.06 o 2.27) 1.56 (1.06 o 2.30) 1.50 (0.99 o 2.28) 1.54 (1.01 o 2.35) S eng hening, n=126 1.95 (1.29 o 2.93) 1.77 (1.17 o 2.69) 1.68 (1.07 o 2.66) 1.57 (0.99 o 2.49) Poo , n=130 3.14 (2.10 o 4.70) 2.77 (1.83 o 4.19) 2.87 (1.83 o 4.49) 2.52 (1.59 o 3.98) p Value o LMA <0.001 <0.001 <0.001 0.001 Non-op imal sel - a ed heal h as ou come Pe manen , n=550 1 1 1 1 High le el, n=163 1.38 (0.96 o 2.00) 1.46 (1.00 o 2.13) 1.19 (0.81 o 1.76) 1.26 (0.85 o 1.87) S eng hening, n=126 1.57 (1.05 o 2.36) 1.52 (1.00 o 2.29) 1.44 (0.94 o 2.20) 1.40 (0.91 o 2.16) Poo , n=130 2.26 (1.53 o 3.34) 2.00 (1.34 o 3.00) 1.83 (1.21 o 2.77) 1.65 (1.08 o 2.53) p Value o LMA <0.001 0.003 0.025 0.087 Model 0 c ude ORs. Model 1 ORs adjus ed o sociodemog aphic a iables (socioeconomic posi ion, gende , ma i al s a us and pa en al s a us). Model 2 ORs adjus ed o heal h s a us a age 30. Model 3 ORs adjus ed o models 1+2. Waene lund A-K, Gus a sson PE, Hamma s öm A, e al.BMJ Open 2014;4:e004053. doi:10.1136/bmjopen-2013-004053 7 Open Access g oup.bmj.com on June 30, 2016 - Published by h p://bmjopen.bmj.com/Downloaded om The s uc u e behind he concep ualisa ion o LMA needs conside a ion. The LMA s uc u e is based on he idea ha he e is a g adien in he labou ma ke , wi h some posi ions being mo e s ongly a ached o he labou ma ke han o he s. We ha e used a modified e sion o A onsson’s co e pe iphe y model 32 o o de he ype o employmen con ac /si ua ion, om he mos a ached o he leas a ached o labou ma ke . Ou s uc u e is, indeed, a simplifica ion o a mo e complex eali y; his should be conside ed when in e - p e ing he esul s. The LMA a iable and socio- economic posi ion a e pa ly o e lapping, as he g oup o en ep eneu s is pa o he SEP classifica ion and he LMA a iable. This o e lap could cause o e -adjus men in he logis ic eg ession analysis. As hese wo a iables pa ly measu e he same phenomenon, adjus ing o socioeconomic posi ion could adjus o pa o he ue e ec o LMA. CONCLUSIONS Policymake s equen ly p omo e flexible employmen con ac s. Howe e , his s udy shows ha hose wi h he mos a ou able bu long- e m high-le el a achmen su e om wo se heal h han hose pe manen ly a ached. Fu he , despi e a de elopmen owa ds a ou - able employmen ci cums ances, heal h disc epancies can endu e, sugges ing a po en ially ‘sca ing’e ec . Finally, hose poo ly a ached display he wo s heal h si ua ion, which is pa ly due o coexis ing bu densome li e ci cums ances. We fi mly sugges ha policymake s should conside he esul s o his s udy and ake espon- sibili y o c ea ing employmen oppo uni ies sui able o main aining a heal hy wo king li e, a he han p o- mo ing employmen con ac s cha ac e ised by poo a achmen o he labou ma ke . Acknowledgemen s The au ho s would like o hank all he pa icipan s in he No he n Swedish Coho . Liudmila Lipiäinen and Tapio Nummi a e hanked o con ibu ing o his pape wi h ega d o he ajec o y analysis. The au ho s also hank Alan C ozie o p o essional language edi ing. Collabo a o s Liudmila Lipiainen; Tapio Nummi. Con ibu o s AH, PEG, PV and A-KW ha e subs an ially con ibu ed o concep ion and design. A-KW and PEG ha e con ibu ed o analysis and in e p e a ion o da a. A-KW, AH, PV and PEG ha e d a ed he a icle and e ised i c i ically o impo an in ellec ual con en . All au ho s ha e gi en hei inal app o al o he e sion o be published. Funding This wo k was suppo ed by Swedish Council o Wo king Li e and Social Resea ch [2006-0950] he Medical Facul y a Umeå Uni e si y and he Academy o Finland [g an numbe 132668 o PV]. The s udy has been inanced by The Swedish Resea ch Council o En i onmen , Ag icul u al Sciences and Spa ial Planning dn 259-2012-37. Compe ing in e es s None. E hics app o al Regional E hical Re iew boa d in Umeå, Sweden. P o enance and pee e iew No commissioned; ex e nally pee e iewed. Da a sha ing s a emen The No he n Swedish Coho is no eely a ailable. Resea che s who in e es ed in collabo a ion should ge in o con ac wi h he P incipal In es iga o , Anne Hamma s öm; anne.hamma s [email p o ec ed] Open Access This is an Open Access a icle dis ibu ed in acco dance wi h he C ea i e Commons A ibu ion Non Comme cial (CC BY-NC 3.0) license, which pe mi s o he s o dis ibu e, emix, adap , build upon his wo k non- comme cially, and license hei de i a i e wo ks on di e en e ms, p o ided he o iginal wo k is p ope ly ci ed and he use is non-comme cial. See: h p:// c ea i ecommons.o g/licenses/by-nc/3.0/ REFERENCES 1. OECD. Labou and Social A ai s. Incidence o empo a y employmen : OECD Online OECD Employmen da abase [ci ed 12 Decembe 2012]. 2012. h p://www.oecd.o g/employmen /da abase 2. Vi anen P, Lipiäinen L, Hamma s öm A, e al. T acks o labou ma ke a achmen in ea ly middle age: a ajec o y analysis o e 12yea s. Ad Li e Cou se Res 2011;16:55–64. 3. Fu åke B, Be glund T. Changes o Employe , Employmen P o ec ion Legisla ion and Labou Ma ke A achmen . 6 h Con e ence in Memo y o P o esso Ma co Biagi. Modena, I aly. 2008. 4. Depa men o T anspo . Da a Sou ces o he App aisal o Regene a ion Impac s: TAG Uni 3.5.13. h p://www.d .go .uk/ web ag/documen s/expe /pd /uni 3.5.13.pd (accessed 18 No 2013). Go e nmen o he Uni ed Kingdom 2011. 5. Dooley D, Fielding J, Le i L. Heal h and unemploymen . Anu Re Publ Heal h 1996;17:449–65. 6. Hamma s öm A. You h unemploymen and ill heal h: esul s om a wo-yea ollow-up s udy [in Swedish, summa y in English]. Solna: Ka olinska Ins i u e , 1986. 7. McKee-Ryan FM, Song ZL, Wanbe g CR, e al. Psychological and physical well-being du ing unemploymen : a me a-analy ic s udy. J Appl Psychol 2005;90:53–76. 8. B enne H, Ahe n W. Sickness absence and ea ly e i emen on heal h g ounds in he cons uc ion indus y in I eland. Occup En i on Med 2000;57:615–20. 9. Mauno S, De Cuype N, Kinnunen U, e al. Wo k cha ac e is ics in long- e m empo a y wo ke s and empo a y- o-pe manen wo ke s: a p ospec i e s udy among Finnish heal h ca e pe sonnel. Econ Ind Democ acy 2012;33:357–77. 10. Waene lund AK, Vi anen P, Hamma s öm A. Is empo a y employmen ela ed o heal h s a us? Analysis o he No he n Swedish Coho . Scand J Public Heal h 2011;39:533–9. 11. Waene lund A-K. Tempo a y employmen and illness. Depa men o public heal h and clinical medicine. Umeå: Umeå Uni e si y, 2013. 12. A onsson G, Dallne M, Lindh T, e al. Flexible pay bu ixed expenses: pe sonal inancial s ain among on-call employees. In J Heal h Se 2005;35:499–28. 13. Pa ke SK, G i in MA, Sp igg CA, e al. E ec o empo a y con ac s on pe cei ed wo k cha ac e is ics and job s ain: a longi udinal s udy. Pe s Psychol 2002;55:689–719. 14. Fe ie JE, Wes e lund H, Vi anen M, e al. Flexible labo ma ke s and employee heal h. Scand J Wo k En i on Heal h 2008:98–10. 15. Kim MH, Kim CY, Pa k JK, e al. Is p eca ious employmen damaging o sel - a ed heal h? Resul s o p opensi y sco e ma ching me hods, using longi udinal da a in Sou h Ko ea. Soc Sci Med 2008;67:1982–94. 16. Vi anen P, Vah e a J, Ki imaki M, e al. Employmen secu i y and heal h. J Epidemiol Communi y Heal h 2002;56:569–74. 17. Vi anen P, Liukkonen V, Vah e a J, e al. Heal h inequali ies in he wo k o ce: he labou ma ke co e-pe iphe y s uc u e. In J Epidemiol 2003;32:1015–21. 18. Ba dasi E, F ancesconi M. The impac o a ypical employmen on indi idual wellbeing: e idence om a panel o B i ish wo ke s. Soc Sci Med 2004;58:1671–88. 19. Vi anen P, Vah e a J, Ki imaki M, e al. Labo ma ke ajec o ies and heal h: a ou -yea ollow-up s udy o ini ially ixed- e m employees. Am J Epidemiol 2005;161:840–6. 20. Waene lund AK, Gus a sson PE, Vi anen P, e al. Is he co e-pe iphe y labou ma ke s uc u e ela ed o pe cei ed heal h? Findings o he No he n Swedish Coho . BMC Public Heal h 2011;11:10. 21. B ough on A, Bile a I, Kullande M. Flexible o ms o wo k: ‘ e y a ypical’con ac ual a angemen s. Eu opean Founda ion o he Imp o emen o Li ing and Wo king Condi ions, 2010. 22. Liukkonen V, Vi anen P, Vah e a J, e al. Employmen ajec o ies and changes in sense o cohe ence. Eu J Public Heal h 2010;20:293–8. 23. Nagin DS. Analyzing de elopmen al ajec o ies: a semipa ame ic, g oup-based app oach. Psychol Me hods 1999;4:139–57. 24. Hamma s öm A, Janle U. Coho p o ile: he No he n Swedish Coho . In J Epidemiol 2012;41:1545–52. 8Waene lund A-K, Gus a sson PE, Hamma s öm A, e al.BMJ Open 2014;4:e004053. doi:10.1136/bmjopen-2013-004053 Open Access g oup.bmj.com on June 30, 2016 - Published by h p://bmjopen.bmj.com/Downloaded om 25. Jones SRG, Riddell WC. Unemploymen and nonemploymen : he e ogenei ies in labo ma ke s a es. Re Econ S a 2006;88:314–23. 26. G zywacz JG, Dooley D. “Good jobs” o “bad jobs”: eplica ed e idence o an employmen con inuum om wo la ge su eys. Soc Sci Med 2003;56:1749–60. 27. Fu åke B, Be glund T. Changes o employe , employmen p o ec ion and labou ma ke a achmen : an analysis o Swedish da a om 1972–1998. Go henbu g: Uni e si y o Go henbu g, 2008. 28. Laux R. Measu ing labou ma ke a achmen using he Labou Fo ce Su ey. Labou Ma ke T ends [1361–4819] 1997;105:407. 29. Rousseau D. ed Psychological con ac s in o ganiza ions: unde s anding w i en and unw i en ag eemen s. Thousand Oaks, CA: Sage, 1995. 30. Meye J, Paunonen S, Gella ly I, e al. O ganiza ional commi men and job pe o mance: i ’s he na u e o commi men ha coun s. J Appl Psychol 1989;74:152–6. 31. G eenhalgh L, Rosenbla Z. Job insecu i y: owa d concep ual cla i y. Acad Manage Re 1984;9:438–48. 32. A onsson G, Gus a sson K, Dallne M. Fo ms o employmen , wo k en i onmen and heal h in o a cen e—pe iphe y pe spec i e. A be e och Hälsa (Wo k and Heal h). S ockholm: Na ional Ins i u e o Wo king Li e, 2000. 33. S a is ics Sweden. Unde sökning om le nads ö hållande [In English: Su ey o li ing condi ions]. S ockholm: S a is ics Sweden, 1980. 34. Hamma s öm A, Vi anen P, Janle U. A e he heal h consequences o empo a y employmen wo se among low educa ed han among high educa ed? Eu J Public Heal h 2011;21:756–61. 35. Sieg is J, Ma mo M. Heal h inequali ies and he psychosocial en i onmen — wo scien i ic challenges. Soc Sci Med 2004;58:1463–73. 36. Black D, Mo is JN, Townsend P, e al.Inequali ies in heal h. Repo o a esea ch wo king g oup. London: Depa men o Heal h and Social Secu i y, 1980. 37. Ma mo M, Allen J, Goldbla P, e al.Fai socie y, heal hy li es: s a egic e iew o heal h inequali ies in England pos -2010 ( he Ma mo e iew). London: UCL, 2010. 38. Lynch J, Kaplan G. Socioeconomic posi ion. In: Be kman LF, Kawachi I. eds Social epidemiology. New Yo k: Ox o d Uni e si y P ess, 2000:13–35. 39. A azcoz L, Benach J, Bo ell C, e al. Social inequali ies in he impac o lexible employmen on di e en domains o psychosocial heal h.J Epidemiol Communi y Heal h 2005;59:761–7. 40. S a is ics Sweden. Socioeconomisk indelning (SEI). Meddelande i samo dnings ågo 1982:4 (Swedish Socioeconomic Classi ica ion). S ockholm: S a is ics Sweden, 1983. 41. Golsch K. Employmen lexibili y in Spain and i s impac on ansi ions o adul hood. Wo k Employ Soc 2003;17:691–718. 42. Wald on I, Weiss CC, Hughes ME. In e ac ing e ec s o mul iple oles on women’s heal h. J Heal h Soc Beha 1998;39:216–36. 43. A onsson G, Gus a sson K, Dallne M. Wo k en i onmen and heal h in di e en ypes o empo a y jobs. Eu J Wo k O gan Psy 2002;11:151–75. 44. Vi es A, Amable M, Fe e M, e al. Employmen p eca iousness and poo men al heal h: e idence om Spain on a new social de e minan o heal h. J En i on Public Heal h 2013;2013: 978656. 45. Mu hén B, Mu hén L. In eg a ing pe son-cen e ed and a iable-cen e ed analyses: g ow h mix u e modeling wi h la en ajec o y classes. Alcohol Clin Exp Res 2000;24:882–91. 46. Tol anen A. La en g ow h mix u e modelling: a simula ion s udy. Depa men o ma hema ics and s a is ics. Finland: Uni e si y o Jy äskylä, 2007. 47. Cockx B, Picchio M. A e sho -li ed jobs s epping s ones o long-las ing jobs? Ox Bull Econ S a 2012;74:646–75. 48. S a is ics Sweden. Women and men in Sweden 2012. Ö eb o: S a is ics Sweden, 2012. 49. Mi owsky J, Ross C. Social causes o psychological dis ess. New Yo k: Aldine De G uy e , 2003. 50. S ully KW. Job loss and heal h in he US labou ma ke . Demog aphy 2009;46:221–46. 51. Ba ley M. Unemploymen and ill heal h—unde s anding he ela ionship. J Epidemiol Communi y Heal h 1994;48:333–7. 52. Fe ie JE. Is job insecu i y ha m ul o heal h? J R Soc Med 2001;94:71–6. 53. Tompa E, Sco -Ma shall H, Dolinschi R, e al. P eca ious employmen expe iences and hei heal h consequences: owa ds a heo e ical amewo k. Wo k 2007;28:209–24. 54. Facey ME, Eakin JM. Con ingen wo k and ill-heal h: concep ualizing he links. Soc Theo y Heal h 2010;8:326–49. 55. Jung T, Wick ama KAS. An in oduc ion o la en class g ow h analysis and g ow h mix u e modeling. Soc Pe s Psychol Compass 2008;2:302–17. 56. Hynes K, Cla kbe g M. Women’semploymen pa e ns du ing ea ly pa en hood: a g oup-based ajec o y analysis. J Ma iage Fam 2005;67:222–39. 57. Gues DE, Isaksson K, De Wi e H, e al.Employmen con ac s, psychological con ac s, and wo ke well-being: an in e na ional s udy. Ox o d: Ox o d Uni e si y P ess, 2010. 58. Kim IH, Mun ane C, Shahidi FV, e al. Wel a e s a es, lexible employmen , and heal h: a c i ical e iew. Heal h Policy 2012;104:99–127. 59. TCO. Addi ional commen om TCO ega ding in ingemen p ocedu e 2007/4835 agains Sweden. S ockholm he Swedish Con ede a ion o p o essional employees. 2013. h p://www. co.se/ Documen s/Addi ionalcommen s.pd 60. Wei o GR, Rosén M. Is pe cei ed ne ousness and anxie y a p edic o o p ema u e mo ali y and se e e mo bidi y? A longi udinal ollow up o he Swedish su ey o li ing condi ions. J Epidemiol Communi y Heal h 2005;59:794–8. 61. Idle EL, Benyamini Y. Sel - a ed heal h and mo ali y: a e iew o wen y-se en communi y s udies. J Heal h Soc Beha 1997;38:21–37. 62. Wä ne yd B, Tho slund M, Ös lin P. The quali y o e ospec i e ques ions abou occupa ional his o y—a compa ison be ween su ey and census da a. Scand J Soc Med 1991;19:7–13. Waene lund A-K, Gus a sson PE, Hamma s öm A, e al.BMJ Open 2014;4:e004053. doi:10.1136/bmjopen-2013-004053 9 Open Access g oup.bmj.com on June 30, 2016 - Published by h p://bmjopen.bmj.com/Downloaded om