JrournalofEpidemiology and Community Health 1993; 47: 14-18
Health and re-employment in a two
year follow
up
of long term unemployed
Bj0rgulf Claussen, Arild Bj0rndal, Peter F Hjort
National Institute of PublicHealth,Unit forHealthServices Research,Oslo, Norway BClaussen ABj0rndal P FHjort Correspondenceto:
DrClaussen,at:University ofOslo, Departmentof Community Medicine, Unit of SocialInsurance Medicine,POBox95 UllevalSykehus,N-0407 Oslo, Norway
Acceptedforpublication July1992
Abstract
Study objective-The aim was to examine re-employment and changes in health during a two year follow up of a represen- tative sample of long term unemployed.
Design-This was acrosssectional study and a two year follow up. Health was measured bypsychometric testing, Hopkins symptom checklist, General health ques- tionnaire, and medical examination. Health related selection to continuous unemploy- mentand recovery by re-employment was estimated by logistic regression with co- variancesdeduced from the labour market theories of human capital and segmented labour market.
Setting-Four municipalities in Gren- land, southern Norway.
Subjects-Participants were a random sample of17 to63 yearold people registered as unemployed for more than 12weeks.
Main results-In thecrosssectional study, the prevalence of depression, anxiety, and somatic illness was from four to 10 times higher than inacontrol group of employed people. In the follow up study, there was considerable health related selectionto re- employment. A psychiatric diagnosis was associated witha
70%
reduction inchances of obtainingajob.Normalperformanceon psychometrictesting showeda twotothree timesincreased chance ofre-employment.Recovery of health following re-employ- ment wasless thanexpectedfromprevious studies.
Conclusions-Health related selectionto longtermunemploymentseems toexplaina substantial part of the excess mental morbidity among
unemployed people.
An increased proportion of the long term unemployed will be vocationally handi- cappedasyears pass,puttinga heavybur- den onsocial services.JEpidemiol Community Health 1993;47: 14-18
Increasing unemployment is a challenge to the health service inmostWestern countries.Excess morbidityandmortality among
unemployed
are welldocumented,'
2 although we do not know much about theexplanation.
There is growing evidence for the view that unemployment damages health,3 even if some
researchersdoubt that lack ofpaidworkpersecan causesickness.4 5 Another
explanation
isthatbad health predisposes to unemployment, eitherby
job loss(SI
in the figure) or by selection tore-employment(S2 in the
figure).
Selection has been less investigated than the causal hypothesis56 and is not found in most investigations.78Even so manyresearchers find it probable that health related selection explains some of the excessmorbidity among the unem-
ployed.1
If this factor is large, it should have implications for the social and health services.Handicapped job seekers will constitute an increasing proportion of the long term unem- ployed, as they are repeatedly scrutinised by personnel mangers over the years. Caring for them will be achallenge ofincreasing significance.
Westudied health and re-employment in a two yearfollowup of a representative sample of long term unemployed in Norway. Our aims were primarilytoassess the selection tore-employment on account of mental distress and of medical diagnoses.Itisscarcely possible to measure S, in the figure, because that would demand baseline health data forhuge numbers of employees before any redundancies are known.6 We think, how- ever, that S2 (selection to continuous unem- ployment) is more important on today's labour market. The reason is that redundancies have reachedamassive scale, while the exit gate from unemployment is narrow and well guarded by personnel managers.
Secondly, we wanted to determine recovery from mental distress following re-employment, which may indicateacausalrelationshipbetween job loss and mental health. Againthe measureis indirect.C,inthefigure would betherealcausal relationship, but this proportion cannot be entangled from S2 without baseline data.
Methods
The study was carried out in the four municipalities of Grenland, a relatively urban district of Norway, partly because the district resembles thecountry,but also because oneofthe authors hadbeen ageneralpractitioner therefor 10 years. Thirty two per cent of the 92 940 inhabitants are workers, compared to 24% in Norwayas awhole.Otherwise the differencesin the labour market structure are
small.9
The unemployment rate in the area increased from266%
when thestudystartedduringthespringof 1988to 5-1% intheautumn of1990,whichwas near the national levels of 2 0 and488%
respectively.
A random sample of a quarter of those registered unemployed for more than 12 weeks was offered a health
examination,
except the"early
pensioners" over 63 years. Forpractical
reasons, we defined the
unemployed
as those registeredatthelabourmarket authorities.This givesa relativelyhealthysample,
becausea pre-Health andre-employmentinlongterm uneployed 15
H
H ~~~~~~~Re-e-n
E-:c;> c-: hr er Lon(
unemploved uriem
A modelofhealth related selectionto unemploymentandof unemploymentas
ofill health. H=healthy;I=ill; SI=selection tojob loss; S2=selection to cl unemployment; C, =healthy unemployedbecome illas a consequenceoflongt
unemployment; C2 sickunemployedwho improve after re-employment.
condition for registering is the ability whatever job is offered. A limit of 12 durationwas setin ordertoavoid thele related "frictional unemployment".
Three hundredand ten persons attel the first examinationin1988,83°% ofthe
Table I Backgroundcharacteristics (percentageormeans) ofthestudygrog Norwegian registered unemployed," allemployed fromGrenland9anda
representative sample oftheoccupational populationinNorwayl1
Unemployed Employed
Grenland All Grenland Na
studygroup Norwegian census san
1988 unemployed 1980a 192
(n=310) (n 56633) (n=41841) (n Age (years)
16-19 6 1 1 9 E
20-29 46 42 25 2(
30-49 37 33 42 32
50-63 12 8 24 2(
64-70 - 7 - -
Gender('Ofemale) 42 40 40 4,
Weeksregistered
0-12 - 61 - -
13-26 52 15 - -
27-52 32 16 - -
53-80 11 5 - -
81- 5 3 - -
Occupation (previous)
Professional,technical 12 11 21 24
Administrative 2 1 5 5
Clerical 8 8 9 12
Sales 10 10 9 1(
Farmers,fishermen 1 2 2 5
Transport,communication 8 7 7 E
Labourers 43 38 33 22
Service 16 12 1 1 12
Unknown - 9 3 2
Socioeconomicstatus
Unskilled workers 43 - 26 1H
Skilled workers 9 - 12 9
Salariedemployeeslow level 23 - 16 18
Salariedemployeesmedium level 22 - 20 29
Salariedemployees highlevel 1 - 7 9
Selfemployedand liberal
professions 1 - 1 3
Agricultureandfishery - - 4 7
Others 2 - 14 7
Social network(index)
Low 38 - - 33
Medium 44 - - 38
High 19 - - 29
Socialsupport(index)
Low 12 - - 15
Medium 61 - - 6C
High 27 - - 25
Education(averageyears) 10-3 - 10 6 iC
Net householdmonthlyincome per
capita (average£) 318 - 360 368
aAgestandardised after the study group, except for the first two variables
They
wererepresentative
of thepopulation
and of allregistered unemployed
inNorway
withrespectto the available variables in table I. The results may be
generalised
to allNorwegian registered long
termunemployed people.'0
Anewexamin- ationtwo yearslaterwasattended by 291 ofthem,giving
anoverallparticipation
rateof780,.
Two hundred and seventy seven answered the tests twiceand this is thetotal numberinmostof the nployed tables. They did not differ from theoriginal sample
ingender (p
=0354),
age(p
=0439),
orH municipality(p=
0O304).
The examination consisted of a structured ...-.-. interview
by
anurse,pen andpaper tests, and the 1 doctor'sdiagnoses and advice. In order to estab- i j lish a reference group of unemployed people, gterr crucialquestions
were taken from national sur- ovec veys. Sociodemographic variables from the last;a cause census gave data
comparable
with allpeople
ontinuous
employed
inGrenland.9
Moredetailed informa-term tion was obtained with variables from the Health
Survey
198512 and the Level ofLiving Survey
to take 1987.13
2 weeks' The index for social network was constructed ss health from sixquestions aboutparents,spouse, child- ren, friends, andfrequencies of visits; the index nded for for socialsupportwasderived from two questions sample. aboutexpectationsof financial and practical help whenneeded.'0 Socioeconomicstatuswasdeter- minedfroma sevenstep indexbasedonoccupa- up,all tion. 13
The
Hopkins
symptom checklist is a 30 item versionof thepsychometric
testSCL-90.14Itwasconstructedfor the Health
Survey
1985andis thetional
only test with a satisfactory Norwegian reference 87apopulation
ofemployed people.
12 A similar=2638)
version is also used in unemployment research.8 The three subscalesintableIIIarequite
likethecorresponding
ones in the much used 58 item6
1version.4 Here,
the cutoffpoint
forpsychiatric
°
morbidity
is )1-75peritem,
which iscommonly
7 used because it gave
prevalences
of2 50o in anAmerican household survey.
The General health
questionnaire (GHQ)
is much used formeasuring psychological
dis-tress,'5
and has beenapplied
to the unem-ployed.7 1-18
We used Likertscoring
as a 4 continuousmeasurementandacutoffpoint
>5in 5 theGHQ scoring
for"casefinding"
withthis 28 2D item version.5 The somatic
diagnoses
in table IIrequired
a2 medical condition that
gives
anenduring
and3
significant
reduction of the common level offunctioning. Psychiatric syndromes
werediag-
3 nosed after the criteria in the
Diagnostic
and statisticalmanualof
mental disorders(DSM-III)
axisI,
andpersonality
disorders after axis II.When
examining re-employment
in relationtohealth,
we assumed that the mostimportant
7
confounding
variables couldbededuced fromthe theories ofhumancapital (education,
yearsinpaid
3
work,
anddurationof thelongest lasting job)
and of segmented labour market (gender, socio- economic status, and socialnetwork).i9
Anotheraspect of labour market conduct is personal involvement in paid work, which was measured
5 ~~~~~~~~~~~~20
8 bytheEmploymentcommitment test.
ThehypothesesweretestedbyX2forcategori-
3 calvariables andtwotailedttestsforcomparisons ofmeans.Logisticandmultiplelinearregressions
I
BjorgulfClaussen,ArildBj0rndal,Peter F Hjort Table II Average scores per item (95q' confidence intervals)for psychometrictests
atthe health examinations in 1988 and 1990
1988 1990
Test Unemployed Employeda Unemployed Re-employed
scales (n=298) (n=4924) (n=164) (n=113)
Hopkins symptom checklist
Somatic symptoms 141 1.19 144 140
(1-38-1 45) (1-16-122) (140-148) (1-36-144)
Anxiety 140 1 13 143 130
(1 35-1 45) (1-12-1 15) (137-1-49) (126-134)
Depression 147 108 147 130
(1 41-1-53) (105-1 13) (1-41-1*53) (126-134)
General health questionnaire 0-88 - 082 069
(082-0{94) (076-0-88) (065-073)
aAhousehold health survey, age standardised after the study
group12
werecarried out bySPSS-PC.21 Covariates were thesevenvariables mentioned above.
Results
THE CROSSSECTIONAL EXAMINATION
Table Ishowsthat thestudygroup was relatively young,whilethe sex ratio was the same as among theemployed.The restofthe tables aretherefore agestandardisedafter the studygroup.
The unemployed had fewer resources like household income (p<0001), socioeconomic status(p<0001), and education(p=<001)than employed people in the Grenland area. Social network was poorer than in thenational sampleof employed (p<00001), social support about the same (p=0223).
The study group scored much higher on the Hopkins symptom checklist than the reference group (table II). The prevalences of mental disorderswerefourtotentimeshigher (tableIII).
These differences were the same for bothgenders
Table III Prevalences of sickness (percentage) afterdichotomous healthmeasure- ments atthe health examinationsin1988 and 1990
1988 1990
Health Unemployed Employed Unemployed Re-employed
measurement (n=298) (n=4924) (n=164) (n=113)
Psychometrictests
Hopkins symptom checklist
Somatic symptoms 17 4t 18 13
Anxiety 17 2t 20 12
Depression 22 21 25 13*
General healthquestionnaire 36 - 34 28
Doctor'sdiagnoses
Somatic disease 26 - 34 23*
Psychiatric syndrome 29 - 41 16t
Personalitydisorder 10 - 13 3t
*p<005;tp<0-01;
p<l0-001
comparedtounemployedatthesametimebyX2testsand did not depend on socioeconomic status, household income, social network, or reported alcohol consumption,asdescribed elsewhere.'0
HEALTH RELATEDSELECTION TO RE-EMPLOYMENT
Fortyonepercent werere-employedwithintwo years (table IV). The differences in education, socioeconomic status, and social network show
TableIV Sociodemographiccharacteristics in1988afteremploymentstatus in1990 Re-employed Unemployed (n=118) (n=173)
Gender(females, ,,) 45 43
Age(averageyears) 32-1 34.3
Socioeconomicstatus(averageindexscore) 2-46 2 17*
Education(averageyears) 10 6 10.1*
Nethousehold income percapita(averageC) 308 331
Social network(averageindexscore) 10-5 9.6*
Social support(averageindexscore) 2 38 236
Employmentcommitmenttest(average score) 431 426
Time inpaidwork(averagemonths) 130 155
Longestlastingjob (averagemonths) 64-5 87.9*
*p<
0.05
byttestthat the moreresourcefulunemployed had better chances of getting a job when education, socioeconomic status, and social network are recognisedas resources. Gender and the Employ- mentcommitment test were not bivariately con- nected with continuous unemployment. The differences intimeof longest employment and in any paidwork disappeared after controlling for age andsocioeconomic status.
Selection to re-employment was estimated by logistic regression with dichotomised health measures as predictors (table V). The anxiety scale, the General health questionnaire, and the psychiatric diagnoses from the first examination showed a significant effect on re-employment.
The chances for getting a job were reduced to 24-470,for the sickunemployed compared to the healthyunemployed. Continuous scores in multi- ple linear regression gave thesame level of signifi- cance for all the test scales.
In thelogistic regressions, variables other than health had no significant effect on continuous unemployment (gender, education, time in paid work,socioeconomic status, social network, or the Employment commitment test).
RECOVERYAFTER RE-EMPLOYMENT
Table VIshowsthe probability of scoring as sick in the follow up controlled by baseline health status,makingthelogisticequation a measure of the change in relative risks ofscoring as sick in the two groups. The prevalence of mental disorders wasreducedby gettinga job.However, only the depression subscale showed significant differ- ences. The re-employed were 45o as likely to experience depression as people who were still unemployed. The continuousmeasuresgave the same levels of significance in multiple linear regression,except for asignificantrecovery in the General health questionnaire(p=003).
Discussion
Do these figures reflect real health? The main instruments are twopsychometrictests, which are much used and well validatedasmeasuresof mental disorders.'4 15 The doctor'sdiagnoses,onthe other hand, are subject to the low reliability of clinical judgements andmust be takenforwhat theyare.
The actual result of selection to re-employment, however, supports the measurements, because health isan integratedpart of the well established humancapital theoryof labour market conduct.'9
CROSS SECTIONAL RESULTS
Psychiatric disorders are far more common among the long term unemployed than in the working population, even in affluent Norway.
This contradicts the view that poverty following lack of work is the main explanation of excess morbidity among the unemployed.5 We found that, as shown before,717 18 relatively well situ- ated groupsof Scandinavianunemployed people experience the same high degree of mental dis- tress astheircounterparts in theUK'6andUSA.8 Thus the model in thefigureforexplainingexcess morbidityis relevant forNorwegianunemployed.
SELECTION
We found a noteworthy selectionto re-employ- ment, both in terms of mental distress and of 16
Health andre-employmentinlongtermuneployed
Table V Selectiontore-employment: results from logistic regression with the probabilityofre-employmentin 1990asdependentvariable and healthmeasuresin 1988 aspredictors, together with gender, socioeconomic status, social network, education, years inpaid work,durationofthelongestlasting job, and the Employment commitmenttest ascovariates
Regression -a
Predictor (O=healthy, 1= sick) coefficient (B) pvalue eB Psychometric tests
Hopkins symptom checklist
Somatic symptoms 0 14 0-69 0-87
Anxiety 101 0 01 0-36
Depression 021 0 52 0 81
General healthquestionnaire 0-76 001 0 47
Doctor's diagnoses
Somatic disease 0 51 0-08 0 60
Psychiatric syndrome 1 20 <0 001 0 30
Personality disorder 1 43 0 01 0 24
aTheodds ratioforbeingre-employedin 1990for the sickunemployedin 1988comparedto thehealthyunemployed.Thosescoringasanxious had forexample a 360o chance of
re-employmentcomparedtothenon-anxious,giventhesameage,education,etc
Table VI Recovery byre-employment: resultsfrom logistic regressionwith the probabilityof beingsick in 1990asdependentvariable andemploymentstatusin 1990 aspredictor, controlledforthesamehealthmeasurein 1988 and thecovariatesfrom table V
Regression _-a
Dependent variable coefficient (B) p value eB
Hopkins symptomchecklist
Somatic symptoms 0-65 0-14 0-52
Anxiety 0 08 0 86 0 93
Depression 0 80 0-04 045
General healthquestionnaire 0-22 0-45 0 80
aThechangefrom 1988to1990inrelativerisk forbeingsick forre-employed compared to stillunemployed.Gettingajobreduces the chance forscoringasdepressedto45', of the chance for those stillgoingunemployed
medical diagnoses. Sick unemployed people in 1988 had between a quarter and a half of the probability for re-employment in 1990 as the healthy ones. The selection was mostly due to psychiatric illness, while the somatic medical diagnoses showed alowselectioneffect.
Surprisingly, human capital variables otherthan health did not predict re-employment in our sample-not even education or work experience.
Thisprobably means that personnel managers are looking for more subtle traits than we were able to measure.Neitherdid wefindthe sexdifferences in unemployment distress that are shown in many European
countries.'
Employed women in Norway seem tobe as much devoted to paid work asmen.Selectionto re-employment was not found in studies fromMichigan,8Finland,7 and the United
Kingdom,'6
the first of these using the Hopkins symptom checklist, the other two the General healthquestionnaire. The explanation is probably thatthey interviewed other groups living under different labour market conditions than our sample. The three samples consisted of blue collar job seekers in the 20 to 50 year age range with unemployment of any duration, while ours was a representative one of long term unemployed peopleaged 17 to 63 years. The British and the Finnish samples were followed in a stable labour market, the US sample during an upgoing busi- nesscycle. The three countries had higher unem- ployment rates than Norway, but none of them were increasing so rapidly. From the time we started ourstudy in1988 until today, the unem- ployment rate has gone up from20/%
to a con- tinental level ofnearly
80h.*"
Oursample
wasprobably
moreheavily
selectedthan the others.RECOVERY
If excess mental distress
disappears
on re-employment,
thisindicates that causation is thedominating explanation. Again, lack of baseline health datarenders exact calculationsimpossible, becausewe cannot be sure thatrecoveryis com- plete for all re-employed people.
Recovery was moderate in our study, in con- trast to thenearlycomplete recoveryin the three other investigations.78 16 Againthe explanation maybe sample differences. Psychometric scores increasemostin the first three months of unem- ployment andthentendtogodown.7 16Thisleads to bigger differences between unemployed and re-employed people in the other three samples with alargeproportionof shorttermunemployed.
Conditions on the labour market may also be important. Many of our re-employed people enteredjobswith lesssecuritythanbefore,which mayexplainthe lowrecoveryrate.Thesevenitem socioeconomicstatusindex isacrudemeasurefor socialmobility,but here360°0 of there-employed crept downwardsand 180° upwards.
IMPLICATIONS FOR THE HEALTH SERVICE
Our findings raise the question ofwhat society should dowith agrowingpopulationofunhealthy andunsuccessfuljobseekers.Over many yearsof lowunemployment,Social InsuranceinNorway hasgrantedagreatnumberofdisabilitypensions.
Thispracticeisnowunderattack,partlybecause of the sickness inducing effect of a disability pension,22andpartlybecause ofexpenditureona growing number ofpensioners. Bothare serious problems that need attention. Society must in some way take care ofthe most unhappy par- ticipants in the labour market, and reforms should notstrikeattheweakest.
This isimportant also from a preventive view.
Sickunemployedpeopleareespecially predisposed tofurther
deterioration.'
Preventivemeasuresfor people who are unemployed in the long term because of poorhealthshould havepriority.Voca- tional rehabilitation, occupational programmes, and sheltered employmentareimportant,aslong as oursocieties donotprovide primary prevention inthisfield,ie, paid work for all.Thisstudywassupported by the Norwegian Research Council for Scienceand the Humanities and by the CountyLabourAuthorityof Telemark.
1 Smith R. Unemployment and health. A disaster and a challenge.Oxford: OxfordUniversity Press,1987.
2Martikainen PT. Unemployment and mortality among Finnish men, 1981-85.BMJ71990; 301: 407-11.
3 Smith R.Unemployment:here wegoagain.BMJ 1991;302:
606-7.
4 Kagan AR. Unemployment causes ill health: the wrong track. Soc Sci Med 1987; 25: 217-8.
5 Bartley M. Unemployment and health: selection or causation-afalse antithesis? Sociol Health Illness1988;10:
41-67.
6 Morris JK, Cook DG. A critical review of the effect of factoryclosuresonhealth. Br JIndMed1991;48:1-8.
7 Lahelma E. Unemployment, re-employment and mental well-being. A panel survey of industrial jobseekers in Finland.ScandjSoc Med 1989;suppl43.
8 Kessler RC, Turner JB, House JS. Unemployment, reemploymentand emotionalfunctioninginacommunity sample.AmSociol Rev1989; 54:648-57.
9Central Bureau of Statistics.Populationandhousingcensus 1980.Kongsvinger:Central Bureau ofStatistics,1982.
10 Claussen B. Arbeidslos og helselos? En medisinsk undersokelse avlengretidsarbeidslosefraGrenland 1988.Oslo:Reportno
2/1991, Unit for Health Services Research, National Institute of Public Health.(Englishsummary)
11Directorate of Labour. Labour MarketStatisticsno 10/1988 and 10/1992.Oslo:Directorate ofLabour,1988and 1991.
12Central Bureau of Statistics. Health survey 1985. Oslo:
Central Bureau ofStatistics,1987.
13 Central Bureau of Statistics. Level of living survey 1987.
Oslo: Central Bureau ofStatistics,1988.
17
Bj0rgulf Claussen, Arild Bjorndal, Peter F Hjort 14Derogatis LR, Lipman RS, Rickels KL, Uhlenhuth EH,
Covi L. The Hopkins Symptom Checklist (HSCL): a measureof primarysymptomdimensions. In: Pichot D, ed.
Psychological measurements inpsycopharmacology. Basel:
KargerVerlag, 1974.
15 GoldbergD, Williams P. A user's guidetothe General health questionnaire. Winsor: NEFR-Nelson Publishing Com-
pany,1988.
16 Warr P, Jackson P. Factors influencing thepsychological impact of prolonged unemployment and ofre-employment.
Psychol Med 1985; 15: 795-807.
17 Iversen L, Sabroe S. Psychological well-being among
unemployed and employed people after a company
closedown: a longitudinal study. Soc Issues 1988; 44:
141-52.
18 Brenner SO, Levi L. Long-term unemployment among womenin Sweden.Soc Sci Med1987; 25: 153-61.
19 Fallon P, Verry D. The economics of labour markets. Oxford:
Philips Allan Publisher Limited, 1988.
20Warr P,Cook J, Wall T. Scales for themeasurementofsome
workattitudes andaspectsof psychological well-being.
Occup Psychol 1979; 52: 129-48.
21 Norusis MJ. SPSS/PC for the IBM PC. Chicago: SPSS Inc., 1986.
22 Westin S. Thestructureofafactory closure: individual
responses to job-loss and unemployment in a 10-year controlledfollow-up study. Soc Sci Med 1990; 31: 1301-11.
18