78
2018 SYSTEMATIC MAPPING REVIEW: Effects of partial sick leave versus full- time sick leave on sickness absence and work participation REPORT

Effekt av gradert sykmelding vs full sykmelding...2 Content Published by Norwegian title Folkehelseinstituttet Område for helsetjenester Effekt av gradert sykmelding vs. full sykmelding

  • Upload
    others

  • View
    6

  • Download
    0

Embed Size (px)

Citation preview

  • 2018SYSTEMATIC MAPPING REVIEW:

    Effects of partial sick leave versus full-time sick leave on sickness absence and work participation

    REPORT

  • 2 Content

    Publishedby

    Norwegiantitle

    FolkehelseinstituttetOmrådeforhelsetjenesterEffektavgradertsykmeldingvs.fullsykmeldingpåsykefraværogarbeidstilknytning:ensystematiskkartleggingsoversikt

    Englishtitle Effectsofpartialsickleaveversusfull‐timesickleaveonsicknessabsenceandworkpar‐ticipation:asystematicmappingreview

    Responsible CamillaStoltenberg,directorAuthors

    JoseF.Meneses‐Echavez,projectleader,researcherNikita,Baiju,researcherRigmorC.Berg,departmentdirector

    ISBN 978‐82‐8082‐917‐7Typeofreport SystematicmappingreviewNo.ofpages 56(77inklusivvedlegg)Comissioner NAV

    Subjectheading(MeSH)

    Sickleave,returntowork,disabilityevaluation

    Citation

    Meneses‐EchavezJF,BaijuN,BergRC.Effectsofpartialsickleaveversusfull‐timesickleaveonsicknessabsenceandworkparticipation:asystematicmappingreview.Report−2018.Oslo:Folkehelseinstituttet,2018.

  • 3 Content

    Content

    CONTENT 3

    HOVEDBUDSKAP 5

    SAMMENDRAG 6

    KEYMESSAGES 9

    EXECUTIVESUMMARY(ENGLISH) 10

    PREFACE 13

    ABBREVIATIONS 14

    BACKGROUND 15

    METHODS 21 Whatisasystematicmappingreview? 21 Selectioncriteria(identifyingtheresearchquestion) 22 Literaturesearch(identifyingrelevantstudies) 23 Studyselection 24 Dataextraction(chartingthedata) 24 Qualityappraisaloftheincludedstudies 24 Collatingandsummarizingtheresults 25 RESULTS 26 Searchresults 26 Descriptionofincludedstudies 27 Qualityappraisaloftheincludedstudies 35 Summaryofmainfindingsfromtheincludedstudies 35 DISCUSSION 45 Mainfindings 45 Generalizabilityandstrengthoffindings 46 Comparisonwithotherreviews 48 Strengthsandweaknesses 48 Implicationsforpractice 49 CONCLUSION 50

    REFERENCES 51

  • 4 Content

    APPENDIX 57 Appendix1.Glossary 57 Appendix2.Searchstrategies 58 Appendix3.Excludedstudies 61 Appendix4.Adjustedanalysesandcovariatesintheregistry‐basedstudies 62 Appendix5.Definitionofthestatisticalanalysesintheregistry‐basedstudies 73 Appendix6.Qualityappraisaloftherandomizedcontrolledtrial 75 Appendix7.Qualityappraisaloftheregistry‐basedstudies 76

  • 5 Hovedbudskap

    Hovedbudskap

    Høytsykefraværblantarbeidstakereerenbekymringimangeland.Gra‐dertsykmeldingerenkombinasjonavarbeidogsykepengersombenyt‐tesnårarbeidstakerenerdelvisarbeidsufør,slikatdenansattekanværefraværendefrajobbendelavtidenogjobbeendelavtiden.Folkehelse‐instituttetfikkioppdragavNAVåutføreenkartleggingavdenempiriskeforskningenomeffektenavgradertsykmeldingversusfullsykmelding.MetodeViutførteensystematiskkartleggingsoversikt.Ijanuar2018gjordevietomfattendelitteratursøk,inkludertsøkistoredatabaser,referanselister,grålitteratur,ogvikontaktetarbeidslivsorganisasjonerogdepartement.Toforskerevurderteuavhengigavhverandrealleidentifisertereferan‐serogdenmetodiskkvalitetentildeinkludertestudiene.Vihentetutdatafradeinkludertestudieneogutførtebeskrivendeanalyser.Synteseavindividuellestudieresultaterinngårikkeiensystematiskkartleg‐gingsoversikt.ResultaterViinkluderteenrandomisertkontrollertstudieog12registerbasertestudier.De13studieneinkluderteca.2,74millionersykmeldte.Studienehaddefølgendekjennetegn: Elleveavstudienevarfranordiskeland,inkludertfirefraNorge. Allestudienehaddeentenmoderatellerhøymetodiskkvalitet. Denrandomisertekontrollertestudieninkludertefinskeansatte

    (n=62)somvarsykmeldtpågrunnavmuskel‐ogskjelettplager,mensderegisterbasertestudienehovedsakeliginkluderteansattemedmuskel‐ogskjelettplagerellerpsykiskelidelser.

    Detvar15utfall,hvoravhovedutfallenevararbeidsdeltakelse,varighetavsykefravær,arbeidsførhetogsosialestønader.

    Funnenetyderpåatgradertsykmeldingerforbundetmedflerepositiveutfall,sliksomhøyerearbeidsdeltakelseogkorteresykefravær,mensikrekonklusjoneromeffekteneavgradertsykmeldingerbegrensetpågrunnavdetstoreflertalletavobservasjonsstudierpådettetemaet.

    Tittel: Effekt av gradert sykmelding vs. full sykmelding på sykefravær og arbeidstilknytning: en systematisk kartleggingsoversikt ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐Publikasjonstype: Systematisk Kartleggingsoversikt En systematisk karleggingsoversikt kartlegger og kategoriserer eksisterende forskning på et tema og identifiserer forskningshull som kan lede til videre forskning. ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐Svarer ikke på alt: - Ingen syntese av resultater - Ingen vurdering av evidensen ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐Hvem står bak denne publika-sjonen? Folkehelseinstituttet har gjennomført oppdraget etter forespørsel fra NAV ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐Når ble litteratursøket utført? Søk etter studier ble avsluttet januar 2018.

  • 6 Sammendrag

    Sammendrag

    Bakgrunn

    Gittdethøyesykefraværblantarbeidstakereerdetimangelandenprioritetåøkear‐beidsdeltakelsenblantpersoneriarbeidsføralder.Gradertsykmeldingerenkombina‐sjonavarbeidogsykepengersombenyttesnårarbeidstakerenerdelvisarbeidsufør,slikatdenansattekanværefraværendefrajobbendelavtidenogjobbeendelavti‐den.Detforståssometgradertfravær,somgjøratfolkmedredusertarbeidskraftkanjobbedeltidogfortsattbeholdetilknytningentilarbeidsmarkedet.Gradertsykmeldingvarierermellom20%og99%,ogkanmuliggjøreraskerereturtilarbeid.Imangeland,spesieltdenordiskelandene,erdettetiltaketsettsomdetførstealternativetforåtakleøkendesykefraværogforhindreekskluderingfraarbeidsmarkedet.Forskningpågradertsykmeldingharhovedsakeligvistlovenderesultater,spesieltidenordiskelandene.EnnyliguførtevalueringavordningenmedgradertsykmeldingiNorgeidentifiserteimidlertidbarrierernårdetgjelderbrukenavordningen,sliksomomfattendebyråkrati,vanskeligåforstå‐ogtilgjengeligtilgangtilinformasjonsamtmangelpåkvalitetskontroll.Selvomlovenderesultatererpublisertharforskningenomeffektenavgradertsykmeldingblittkritisertavmetodiskegrunner,somforeksem‐pelrisikoforutvalgsskjevhetogsvakhetervedmålingavarbeidsdeltakelse.Peridagfinsdetingensystematiskeanalyseravkunnskapsgrunnlagetforeffekteneavgradertsykmelding.Metode

    Vigjennomførteensystematiskkartleggingsoversiktihenholdtilinternasjonalestan‐darder.Ensystematiskkartleggingsoversikt(ogsåkjentsomsystematicscopingreview)erenkunnskapsoppsummeringsomkartleggerogbeskrivereksisterendelitteratur‐ellerforskningsgrunnlagpåetbestemttema.Oppsummeringeninkludereringensyste‐matisksynteseavindividuellestudieresultater(f.eks.ingenmetaanalyse).Viinkludertestudieruansettspråksomevaluerteeffekteneavgradert‐versusfullsykmeldingpåsy‐kefraværogarbeidsdeltakelse.Relevantestudiedesignvarprospektivekontrollertestudiersamtregisterbasertestudier(RB).Prosjektteamet(forskerne)ogoppdragsgive‐ren(NAV)diskuterteogbleenigeomprosjektplanen.Visøkteiåttestorelitteraturdatabaser(frabegynnelseninntil2018),kontakteteksper‐ter,ogsøktewebsidenetilarbeidslivsorganisasjonerogrelevantedepartement,refe‐ranselistenetilalleinkludertestudieroglitteraturoversikterpåtemaet.Toforskerevurderteuavhengigavhverandredeidentifisertereferanseneogdatauthentingenble

  • 7 Sammendrag

    dobbeltsjekket.Vigjennomførteuavhengigkvalitetsvurderingavdeinkludertestudi‐enemedbrukavvalidertesjekklister.Vigruppertedatahentetfradeinkludertestudi‐eneihenholdtildereshovedkarakteristika,utførtebeskrivendeanalyserogpresent‐erteresultateneitekstogtabeller.Resultat

    Trettenstudier,publisertmellom2010‐2017,møtteinklusjonskriteriene.Allestudienemålteeffekteravgradertsykmeldingsammenlignetmedfullsykmeldingblantvoksnesykmeldte.Viinkluderteenfinskrandomisertkontrollertstudie(RCT)(n=62somvarsykmeldtpågrunnavmuskel‐ogskjelettplager)og12RBer(n=2,742,497somvarsyk‐meldtpågrunnavhovedsakeligmuskel‐ogskjelettplagerellerpsykiskelidelser).Del‐takerneidetolvRBenevarfraNorge,Danmark,Finland,SverigeogTyskland.Detvarulikegraderavgradertsykmeldingistudiene.IdenfinskeRCTenble50%gra‐dertsykmeldinggitttil70%avdesykmeldte,mens30%avdesykmeldtearbeidetkorteretimer3‐4dageriuken.IRBenevar50%denhyppigstbruktegraderingen.RBeneanalysertelandsdekkenderegisterdatapåsykefravær(bådegradert‐ogfullsyk‐melding)mellom2001og2014.MedhensyntilstudienesmetodiskekvalitethaddedeninkluderteRCTenmoderatkvalitet;dethaddeogså11avde12RBenemensenRBhaddehøymetodiskkvalitet.Deterviktigåfremheveatregisterbasertestudiererdår‐ligegnettilåpåvisekausalesammenhenger.SammendragavhovedfunnfradeinkludertestudieneSykefravær:DenfinskeRCTenogtreRBerrapportertepositiveresultateravgradertsykmelding,sammenlignetmedfullsykmelding,påsykefraværvedettårsoppfølging.ArbeidsdeltakelseReturn‐to‐work(RTW):DenfinskeRCTenfantatgradertsykmeldingforbedretar‐beidsdeltakelsensammenlignetmedfullsykmeldingvedettårsoppfølging.AlledeniRBenesommåltearbeidsdeltakelse,bortsettfraennorskRB,vistebedreresultaterblantdemedgradertsykmeldingsammenlignetmedfullsykmelding.Arbeidsledighet:TreRBer–fraNorge,TysklandogFinland–rapportertelaverear‐beidsledighetblantpersonersomhaddeværtpågradertsykmeldingsammenlignetmedpersonersomhaddeværtfulltidssykmeldt.

    GradavuførhetogattføringGjentakendesykefravær:DenfinskeRCTenfantingensignifikanteforskjellermellomgruppensomhaddegradertsykmeldingogdesomhaddefullsykmeldingnårdetgjel‐dergjentakendesykefravær.Funksjonsnivå:Ingenforskjellermellomgradert‐ogfullsykmeldingblerapportertidenfinskeRCTenvedettårsoppfølgingogiénRBfraNorge,menstoandreRBer(fraNorgeogFinland)fantatgradertsykmeldingvarforbundetmedbedringerifunksjons‐nivå.

  • 8 Sammendrag

    Produktivitetstap:KundenfinskeRCTenrapportertepåutfalletproduktivitetstap.DatafradenneRCTenvisteatdetikkevarsignifikantforskjellpåproduktivitetstapmellomgradert‐ogfullsykmeldingvedettårsoppfølging.Uførepensjon:EnnorskRBvisteatgradertsykmeldingvarforbundetmedenhøyeregradavuførepensjonsammenlignetmedfullsykmelding.DetotyskeRBenerappor‐terteenredusertrisikoforåmottauførepensjonhosansattepågradertsykmelding.EnfinskRBfantatgradertsykmeldingvarforbundetmedlavererisikoforfulluførepen‐sjonsammenlignetmedfullsykmelding,mensmotsattsammenhengblefunnetforri‐sikofordelvisuførepensjon.Sosialestønader:FireRBerfantatgradertsykmeldingvarforbundetmedenlaveregradavsosialestønadersammenlignetmedfullsykmelding.HelserelaterteutfallBaretoavdeinkludertestudienerapportertepåhelserelaterteutfall.DenfinskeRCTenfantingenforskjellermellomgradertogfullsykmeldingpåsmerteintensitet,menvistepositiveresultaterforgradertsykmeldingpåbådeselvrapportertgenerellhelseoghel‐serelatertlivskvalitet.EntyskRBvisteatpersonermedgradertsykmeldingfungertebedrefysiskogfølelsesmessigenndesomhaddefullsykmelding.DenfinskeRCTenfantingenforskjellermellomgradertogfullsykmeldingnårdetgjaldtdesykmeldtesdepresjonssymptomer,mensdentyskeRBenvisteatgradertsykmeldingvarforbundetmedforbedringermedhensyntildepresjon‐ogangstsymptomerogarbeidsevne,sam‐menlignetmedfullsykmelding.Konklusjon

    Forskningsgrunnlagetforeffektenavgradertsykmeldingsammenlignetmedfullsyk‐meldingbeståravénRCTog12RBer,medtotaltca2,74millionerpersonersomersyk‐meldtpågrunnavhovedsakeligmuskel‐ogskjelettplagerellermentalelidelser.Deink‐ludertestudienevistesammenfallendemønsteriresultatifavøravgradertsykmelding.BådeRCTenogRBeneindikerteatgradertsykmeldingerforbundetmedkorteresyk‐meldingoghøyerearbeidsdeltakelse.ResultatenefradenfinskeRCTenindikerteatan‐sattemedmuskel‐ogskjelettsykdommerrapportertebedregenerellhelseoglivskvali‐tetmedgradertsykmelding.DenneRCTenfantatgradertsykmeldingikkeharnoenef‐fektpågjentakendesykefravær,produktivitetstapellersmerte.Positivesammen‐hengermellomgradertsykmeldingogforbedringavansattesfunksjonshemmingogdepressivesymptomerblestøttetavRBene,menikkeavRCTen.ResultaterfraRBeneantyderatbrukavgradertsykmeldingerforbundetmedlaveresannsynlighetforåmottabådeuførepensjonogsosialestønader,samtbedreskårepåfysisk‐ogfølelses‐messigfungering,angstogarbeidsevne.SikrekonklusjoneromeffekteneavgradertsykmeldingerbegrensetpågrunnavdetstoreflertalletavRBerpådettetemaet.ObservasjonellestudiersomRBergirenbety‐deligrisikoforsystematiskeskjevheteriresultatenesomgjørdetvanskeligågisikresvarpåspørsmålomårsakogvirkning.FlereRCTeravhøymetodiskkvaliteternød‐vendigforåkunnetrekkeklarekonklusjoner.

  • 9 Keymessages

    Keymessages

    Inmanycountries,thehighsicknessabsencerateinworkingagepeopleisaconcern.Partialsickleave(PTSL)isareturn‐to‐workstrategythatenablesemployeestobeabsentfromworkpartofthetimeandremainworkingforaproportionofthetime.TheNorwegianLaborandWelfareAdministration(NAV)commissionedtheNorwegianInstituteofPublicHealthtomapallevidenceontheeffectsofPTSLversusfull‐timesickleave(FTSL)onsicknessabsenceandworkparticipation.MethodsWeconductedasystematicmappingreview.InJanuary2018,wecon‐ductedanextensiveliteraturesearch,includingsearchesinmajordata‐bases,referencelists,greyliterature,andwecontactedlaboragenciesandinternationalministries.Twoindependentreviewersscreenedallre‐trievedrecordsandappraisedtheincludedstudies.Weextracteddatafromtheincludedstudiesandperformeddescriptiveanalyses.Synthesisofindividualstudyresultsisnotpartofsystematicmappingreviews.ResultsWeincludedonesmallrandomizedcontrolledtrialand12registry‐basedstudies.The13studiesincludedabout2.74millionemployeesonsickleave.Thestudiesexhibitedthefollowingcharacteristics: ElevenofthestudieswerefromNordiccountries,includingfourfrom

    Norway. Allstudieshadeithermoderateorhighmethodologicalquality. TherandomizedcontrolledtrialincludedFinnishemployees(n=62)

    whoweresick‐listedduetomusculoskeletaldisorders,whiletheregistry‐basedstudiesmostlyincludedemployeeswitheithermusculoskeletal‐ormentaldisorders.

    Therewere15outcomes,ofwhichthemostfrequentlyreportedoutcomeswereworkparticipation,sicknessabsenceduration,disability,andsocialwelfarebenefits.

    ThefindingsindicatedthatPTSLmaybeassociatedwithseveralfavora‐bleoutcomessuchasshortersicknessabsenceandhigherworkpartici‐pation.However,firmconclusionsabouttheeffectsofPTSLcannotbedrawnduetotheoverwhelmingmajorityofobservationalstudiesinthisbodyofevidence.

    Title: Effects of partial sick leave ver-sus full-time sick leave on sick-ness absence and work partici-pation: a systematic mapping re-view ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐Type of publication: Systematic mapping re-view A systematic mapping review maps out and categorizes exist-ing research on a topic, identify-ing research gaps that can guide future research. ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐Doesn’t answer everything: No synthesis of the results No recommendations are made ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐Publisher: Norwegian Institute of Public Health ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐Updated: Last search for studies: January 2018.

  • 10 Executive summary (English)

    Executivesummary(English)

    BackgroundInmanycountries,highsicknessabsenceratesmeanthattheneedtoincreaseworkparticipationofworkingagepeopleisacriticalpriority.Partialsickleave(PTSL),alsocalledgradedsickleave,isareturn‐to‐workstrategythatenablesemployeestobeab‐sentfromworkpartofthetimeandremainworkingforaproportionofthetime.Itisunderstoodasagradedleave,whichallowspeoplewithreducedworkabilitytoworkparttimeandstillkeepthelinktothelabormarket.PTSLvariesbetween20%upto99%,andmightfacilitateaprogressivereturntowork.Inmanycountries,especiallytheNordiccountries,thisreturn‐to‐workstrategyhasbeenconsideredthefirstoptiontotackleincreasingsicknessabsenceratesandtopreventlabormarketexclusion.ResearchonPTSLhasshownlargelypromisingresults,especiallyintheNordiccoun‐tries.ArecentevaluationofthePTSLarrangementinNorway,however,highlightedbarrierstoitsuse,suchasbureaucraticredtape,lackofeasytounderstandandaccessinformation,andqualitycontrol.Althoughpromisingresultshavebeenpublished,re‐searchontheeffectsofPTSLhasbeencriticizedonmethodologicalgrounds,suchasriskofselectionbiasandweaknessesinmeasuringworkparticipation.Todate,nosys‐tematicanalysisoftheevidencebaseontheeffectsofPTSLhasbeenundertaken.ObjectiveTheNorwegianLaborandWelfareAdministration(NAV)commissionedtheNorwegianInstituteofPublicHealthtomapallevidenceontheeffectsofpartialsickleave(PLSL)versusfull‐timesickleave(FTSL)onsicknessabsenceandworkparticipation.MethodWeconductedasystematicmappingreviewaccordingtointernationalstandards.Asystematicmappingreview(alsoknownassystematicscopingreview)isareviewthatmapsanddescribestheexistingliteratureorevidencebaseonaparticulartopic.Thereisnosystematicsynthesisofindividualstudyresults(e.g.,nometa‐analysis).Wein‐cludedstudiesinanylanguageevaluatingtheeffectsofPTSLversusFTSLonsicknessabsenceandworkparticipation.Eligiblestudydesignswereprospectivecontrolledstudiesaswellasregistry‐basedstudies(RBs).Theprojectteam(reviewers)andcom‐missioner(NAV)discussedandagreedontheresearchprotocol.Wesearchedeightmajordatabases,frominceptionto2018,contactedexperts,andhandsearchedwebsitesoflabororganizationsandministries,thebibliographiesofallincludedstudies,andliteraturereviewsinthefield.Twoindependentreviewers

  • 11 Executive summary (English)

    screenedtheretrievedreferencesanddataextractionwasdouble‐checked.Wecon‐ductedindependentqualityappraisaloftheincludedstudiesbyusingvalidatedcheck‐lists.Wegroupeddataextractedfromtheincludedstudiesaccordingtotheirchiefcharacteristics,performeddescriptiveanalyses,andpresentedtheresultsintextandtables.ResultsThirteenstudies,publishedbetween2010‐2017,metourinclusioncriteria.Allevalu‐atedtheeffectsofPTSLcomparedtoFTSLinsick‐listedemployees.WeincludedoneFinnishrandomizedcontrolledtrial(RCT)(n=62sick‐listedemployeesduetomusculo‐skeletaldisorders),and12RBs(n=2,742,497sick‐listedemployeesduetomainlymus‐culoskeletal‐ormentaldisorders).Theparticipantsinthe12RBswerefromNorway,Denmark,Finland,Sweden,andGermany.DifferentgradesofPTSLwereobservedacrossthestudies.IntheFinnishRCT,a50%PTSLwasgivento70%ofallsick‐listedemployees,whereas30%ofemployeesworkedshorterhourson3–4daysperweek.IntheRBs,themostcommonlyusedPTSLwas50%.TheRBsanalyzednationwideregistrydataonsicknessabsences(bothPTSLandFTSL)grantedbetween2001and2014,whichindicateda13‐years’timeline.Regard‐ingthestudies’methodologicalquality,theincludedRCThadmoderatequality,asdid11ofthe12RBswhileoneRBhadhighmethodologicalquality.ItisimportanttonotethatRBsdonotenableresearcherstoestablishcausalrelationshipsbetweenaninter‐ventionorexposureandoutcomes.SummaryofmainfindingsfromtheincludedstudiesSicknessabsence:TheFinnishRCTandthreeRBsreportedpositiveresultsofPTSLcom‐paredtoFTSLonsicknessabsenceatone‐yearfollow‐up.Workparticipation:Return‐to‐work:TheFinnishRCTfoundthatPTSLimprovedworkparticipationcom‐paredtoFTSLattheendoftheone‐yearfollow‐upperiod.AllofthenineRBsthatmeasuredworkparticipation,exceptoneNorwegianRB,showedfavorableassociationsinemployeesonPTSLcomparedtoFTSL.Unemployment:ThreeRBs–fromNorway,GermanyandFinland–reportedfavorableeffectsofPTSLcomparedtoFTSLonemployees’unemployment.Degreeofdisabilityandrehabilitationbenefits:Recurrenceofsickleaveforanycause:TheFinnishRCTfoundnosignificantdiffer‐encesbetweenPTSLandFTSLintherecurrenceofsickleaveforanycause.Disability:NodifferencesbetweenPTSLandFTSLwerereportedbytheFinnishRCTatone‐yearfollow‐upandinoneNorwegianRB,whereastwootherRBs(fromNorwayandFinland)foundthatPTSLwasassociatedwithimprovementsonemployees’disa‐bility.

  • 12 Executive summary (English)

    Productivityloss:OnlytheFinnishRCTreportedonproductivityloss.DatafromthisRCTshowedtherewasnosignificantdifferenceonproductivitylossbetweenPTSLandFTSLuptoone‐yearfollow‐up.Disabilitypension:OneNorwegianRBreportedthatPTSLwasassociatedwithahigherrateofreceivingdisabilitypensioncomparedtoFTSL,whereasthetwoGermanRBsre‐portedadecreasedriskofreceivingdisabilitypensioninemployeesonPTSL.AFinnishRBfoundthatPTSLwasassociatedwithalowerriskoffulldisabilitypensioncomparedtoFTSL,whereastheoppositeassociationwasfoundfortheriskofpartialdisabilitypension.Allowanceofsocialwelfarebenefits:FourRBsobservedthatPTSLwasassociatedwithalowerallowanceofsocialwelfarebenefitscomparedtoFTSL.Health‐relatedoutcomesOnlytwooftheincludedstudiesreportedonhealth‐relatedoutcomes.TheFinnishRCTfoundnodifferencesbetweenPTSLandFTSLonpainintensity,butshowedpositiveresultsforPTSLonbothself‐ratedgeneralhealthandhealth‐relatedqualityoflife.AGermanRBshowedthatPTSLwasassociatedwithbetterphysicalandemotionalfunctioninginsick‐listedemployees.TheFinnishRCTfoundnodifferencebetweenPTSLandFTSLonsick‐listedemployees’depressivesymptoms,whiletheGer‐manRBshowedthatPTSLwasassociatedwithimprovementsonbothdepressionandanxietysymptoms,andworkingability,whencomparedtoFTSL.ConclusionTheevidenceontheeffectsofPTSLcomparedtoFTSLconsistsofonesmallRCTand12RBs,withatotalofabout2.74millionstudyparticipantswithmostlymusculoskeletal‐ormentaldisorders.ThefindingsindicatePTSLmaybeassociatedwithseveralfavora‐bleoutcomes,suchashigherworkparticipation,butduetotheoverwhelmingmajorityofobservationalstudiesinthisbodyofevidence,firmconclusionsabouttheeffectsofPTSLcannotbedrawn.BothstudydesignssuggestedPTSLmaybeassociatedwithshortersicknessabsencedurationandhigherworkparticipation.TheFinnishRCTre‐portedthatemployeeswithPTSLexperiencedbettergeneralhealthandqualityoflifecomparedtothoseonFTSL.However,itdidnotfindstatisticaldifferencesbetweenPTSLandFTSLonsickleaverecurrence,employees’productivityloss,andpain.TheRBsindicatedalowerprobabilityforpeopleonPTSLofreceivingbothdisabilitypen‐sionandallowancebenefits,disability,aswellasbetterscoresonphysical‐andemo‐tionalfunctioning,anxiety,depression,andworkingability.FirmconclusionsabouttheeffectsofPTSLareconstrainedduetotheoverwhelmingmajorityofRBsinthisbodyofevidence.Observationaldesignshaveconsiderableriskofsystematicdifferencesintheresults,whichmakeitdifficulttoanswerquestionsaboutcauseandeffect.FurtherhighqualityRCTsarenecessaryinordertodrawfirmconclusions.

  • 13 Preface

    Preface

    TheNorwegianLaborandWelfareAdministration(inNorwegian:Arbeids‐ogvelferdsetaten,NAV)commissionedtheNorwegianInstituteofPublicHealth(NIPH)tomapoutallevidenceontheeffectsofpartialsickleavecomparedtofull‐timesickleaveonsicknessabsenceandworkparticipation.ThecommissionispartoftheongoingframeworkagreementbetweenNIPHandNAV.ThissystematicmappingreviewcanhelptoinformandsupportNAV,otherauthoritiesaswellasotherresearchorganizationsinevidence‐informeddeliberationsabouttheuseofpartialsickleaveforsick‐listedemployees.TheDivisionforhealthserviceswithintheNorwegianInstituteofPublicHealthfollowsastandardapproachinconductingsystematicmappingreviews,whichisdescribedintheNIPHhandbookSlikoppsummererviforskning.Wemayusestandardformulationswhenwedescribethemethods,resultsanddiscussionofthefindings.Contributorstotheproject:

    Projectcoordinator:JoseMeneses‐Echavez,researcher,NIPH Othercontributors:RigmorCBerg,Departmentdirector;NikitaBaiju,researcher,

    andresearchlibrarianElisabetHafstad,allfromNIPH.Declaredconflictsofinterest:Allauthorsfilledoutaformtodocumentpotentialconflictsofinterest.Noconflictsofinterestweredeclared.TheauthorswouldliketothankElisabetHafstad,researchlibrarianatNIPH,forhelpwithdevelopingthesearchstrategyandforrunningthesearchesinthemajordata‐bases,aswellasHeidNøklebyandDorisToveKristoffersen,bothseniorresearcherswiththeNIPH,fortheirassistancewithcheckingdataaccuracyandguidanceonstatis‐ticalanalysesintheprimarystudies.Wearegratefulforexpertpeerreviewbyprofes‐sorsArneMastekaasa,UniversityofOslo,andPiaSvedberg,KarolinskaInstitute.NIPHisresponsibleforthecontentsofthereviewpresentedinthisreport.Thecom‐missionerofthereviewandpeerreviewersbearnoresponsibility.

    KåreBirgerHagenResearchdirector

    RigmorCBergDepartmentdirector

    JoseF.Meneses‐EchavezProjectleader

  • 14 Abbreviations

    Abbreviations

    CBA,Controlledbefore‐afterstudyCI,ConfidenceintervalCochraneEPOCgroup,TheEffectivePracticeandOrganisationofCareGroupFRW,Fasterreturn‐to‐workFTSL,Full‐timesickleaveGNI,GrossnationalincomeGP,GeneralpractitionerHR,HazardratioIA‐avtalen,InkluderendearbeidslivavtalenITS,InterruptedtimeseriesMD,MentaldisorderMSD,MusculoskeletaldisorderNAV,theNorwegianLaborandWelfareAdministrationNIPH,NorwegianInstituteofPublicHealthNRCT,Non‐randomizedcontrolledstudy/Quasi‐randomizedstudyOECD,OrganisationforEconomicCo‐operationandDevelopmentOR,OddsratioPICO,Population,Intervention,Comparison,OutcomePRISMA‐ScR,PreferredReportingItemsforSystematicreviewsandMeta‐Analysesex‐tensionforScopingReviewsPTSL,Part‐timesickleaveRB,Registry‐basedstudyRCT,RandomizedcontrolledtrialRR,RiskratioRTW,Return‐to‐workSINTEF,StiftelsenforindustriellogtekniskforskningSLS,SemiparametricLeastSquaresanalysis

  • 15 Background

    Background

    Sicknessabsenceandtheneedtoincreaseworkparticipationinworkingagepeoplearecriticalprioritiesindevelopedcountries(1).Notonlydoesahighsicknessabsenceratehavesubstantialcostimplicationsforemployersandsocialsecuritysystems,sick‐nessabsenceaffectsworkers’well‐being,self‐esteemandlivelihood(2).Long‐termsicknessabsenceratesarehighinmanyOECDcountries(2).Norwayexhibitsacom‐parativelyhighrateofsicknessabsence(3),whichhasremainedalmostunchangedforthepastfiveyears(4;5).ThemostrecentdatapublishedbyStatisticsNorway(inNor‐wegian:Statistisksentralbyrå)indicatethesicknessabsencerateinthecountrywas6.5%adjustedforseasonalandinfluenzavariations(self‐anddoctorcertified)inthe3rdquarterof2017.Thisrateisconsiderablyhigherinfemales(8.3%)thaninmales(4.9%)(6).TheNordiccountrieshavesimilarsocialsecuritysystems,andingeneralterms,theyarerelativelysimilarintheirdataonsicknessabsence.However,comparisonsacrossthesecountriesareconstrainedduetodifferencesintheregistrationmethodsinthenationalregisters.Forexample,Swedendoesnotregistersicknessabsencesbelow15days(7).Overall,however,datafromtheNordicSocialStatisticalCommittee2015indi‐catethatNorwayandSwedenhavehighratesoflong‐termsicknessabsence,whereasshort‐termsicknessabsence(i.e.sicknessabsenceoflessthan8days)ishighinDen‐markandlowinNorway(7).Further,thereportstates“womenhavemoresicknessab‐sencethanmen.Olderemployeeshavemorelong‐termsicknessabsencethanyoungeremployees.Youngeremployeeshavemoreshort‐termsicknessabsencethanolderem‐ployees”(7).Togatherthemostup‐to‐datedataonsicknessabsenceinthefiveNordiccountries,weconsultednationalministriesandrelevantorganizations(table1).Briefly,Norwayex‐hibitsthehighestrateofsicknessabsenceintheregion.NorwayandDenmarkpaythehighestpercentageofsalarytotheirsick‐listedemployees(percentageofsicknessben‐efitcomparedtosalary),withthelowestpaymentbeingobservedinFinland.Thehigh‐estnumberofdaysawayfromworkbeforeamedicalcertificatemustbepresentedisseeninSwedenandFinland;whereasDenmarkandIcelandshowthehighestnumberofdaysinwhichtheemployerisresponsibleforpayingsicknessbenefits.

  • 16 Background

    Table1.DataonsicknessabsenceacrosstheNordiccountriesSickleavedata Norway Denmark1 Finland3 Iceland4 Sweden5

    Sicknessabsence 6.5%inthe3rdquarterof2017

    3.7%in2014‐2015

    Ca5%in2015 1‐2%in2006 2.9%inthe3rdquarterof2017

    Numberofdaysawayfromworkbeforeamedicalcertificatemustbepresented

    3days 2days 0‐7days

    3days 7days

    Forhowlongcantheemployeebeoffsickbeforelosingsick‐ness‐benefits?

    260days(52weeks)

    22weekswithin9months.Bene‐fitscanbepaidforalongerpe‐riodifthere‐cipientfulfilstheprolonga‐tioncondi‐tions1

    300days(ex‐cludingSun‐days)overa2‐yearperiod(forthesameillness).120daysforpartialsickness(ex‐cludingSun‐days)overa2‐yearperiod2.

    52weeksovera2‐yearpe‐riod.Foreachmonthworked,anemployeeisentitledtowagesfortwodaysofsickleave.Thegen‐eralruleisthatthesickleaverightsofem‐ployeesin‐creasethelongertheyworkforthesameem‐ployee.

    From1Febru‐ary2016,thetimelimitwasabolishedinthesicknessbenefitsystem.Hence,thereisnolongeramaximumtimeforhowlonganemployeecanbeoffsick.

    Percentageofsick‐nessbenefitcom‐paredtosalary

    100% 90‐100%Sicknesscashbenefit(syged‐agpenge)cal‐culatedbasedonthehourlywageoftheemployee,withamaximumofDKK4,245(€571)perweekorDKK114.73(€15.43)perhour(37hoursperweek),andonthenumberofhoursofwork2

    Ca70%

    Asetfigureir‐respectiveofsalary.Today,governmentpaysIKR1746perday,plusIKR480pereachchildsup‐portedbythepersonreceiv‐ingthebenefits(theamountisdecidedbyministryregu‐lation).

    Ca80%

    Numberofdaysbe‐forebenefitstartsbeingpaid

    None None None None 1

    Numberofdaystheemployerisrespon‐sibleforpayingsick‐nessbenefit

    16days 30days2 10workingdays(includingSaturdays)

    1month 14days

    1. DanishdataprovidedbytheDanishAgencyforLaborMarketandRecruitment(STAR)viaemailcommunication(12/02/18).

    2. DatafromMISSOC,the“MutualInformationSystemonSocialProtection”.3. FinnishdataconfirmedbyEiraViikari‐Junturaandcolleaguesviaemailcommunication(07/03/18).4. IcelandicdatawerepartiallyconfirmedbytheIcelandicHealthInsuranceviaemailcommunication(15/03/18).5. SwedishdataconfirmedbytheMinistryofHealthandSocialAffairsviaemailcommunication(12/02/18).InNorway,employeesareentitledtosicknessbenefitsfromdayoneiftheyhavebeeninpaidworkforthelastfourweeksbeforethesicknesscasestarts(8).Theemployeris

  • 17 Background

    responsibleforpayingthefirst16daysofleave,andthereafterNAVassumesthepay‐ments.Theemployerdesignsafollow‐upplanwithinthefirst4weeks,andameetingwiththesick‐listedemployee,NAV,andthegeneralpractitioner(GP)isarrangedtodis‐cusssolutionsforreturntowork,ensuredialogueandupdatethefollow‐upplan.Asimilarmeetingisalsoorganizedat26weeksofsickleave.Afteroneyearonsickleave,theemployeemaybegrantedatransitionbenefitforuptofouryears(8;9).AlthoughtheapproachestoreducesicknessabsenceintheNordiccountriesvaryduetostructuraldifferencesandvariationsinsicknesspolicies,thecountriesarerelativelycomparablewithregardtotheirpoliciesrelatedtoitandmeasurestoreducesicknessabsence,ascomparedtomanyothercountries,usinginitiativesthatarebasedonclosefollow‐upofthesick‐listedperson.However,alackofevidenceregardingtheeffectsofthosemeasureshasbeenreportedintheliterature(10;11).SomeoftheNordicmeasurestoreducesicknessabsenceanddisabilitypensionincludethefollowing:re‐turntoworkprogram(Denmark),chartingofthepossibilitiesofthelong‐termsicklistedemployees’returntowork(Finland),rehabilitationchain(Sweden),andtheuseofpartialsickleave(possibleinallNordiccountries)(7).Partialsickleave(PTSL),alsocalledgradedsickleave,hasbeenconsideredthefirstop‐tiontotackleincreasingsicknessabsenceratesandtopreventlabormarketexclusion,especiallyintheNordiccountries(3).Accordingtoarecentreport(7),PTSLisunder‐stoodasagradedleave,whichallowspeoplewithreducedworkabilitytoworkparttimeandkeepthelinktothelabormarket.Thatis,itenablesemployeestoreturntomodifieddutiesattheworkplace,andmightfacilitateaprogressivereturntowork(7;12).Anunderlyingpremiseistheexpectationthatworking,andstayingconnectedtotheworkplace,initselfwillcontributetoreducedsickleaveduration(13).PTSLvar‐iesbetween20%upto99%,independentoftheproportionofemployment(8).InNor‐way,useofPTSLgainedtractionafterthe2004restructuringofthenationalsickleaveregulations.Inmostcases,wheneveranemployedpersonaskedforsickleave,theGPshouldconsiderPTSLthedefaultoption(13).AsimilarsituationdevelopedinSwedenafteradvicebythenationalgovernmentintheearly2000stoincreaseuseofPTSL.In2016,PTSLrepresentedaround34%ofallsickleavecasesamongwomenand26%ofthecasesamongmen(14).InNorway,theaimtoreduceratesofsickleavethroughimprovedinclusionmecha‐nismssuchasPTSLisanchoredinthecollectiveagreementoninclusiveworkingcondi‐tions(InNorwegian:Inkluderendearbeidsliv,IA‐avtalen).IA‐avtalenisacollaborativeagreementbetweentheNorwegiangovernmentandsocialpartners.Theagreement,firstsignedin2001,hasbeenrenewedseveraltimes,mostrecentlyin2014forthepe‐riod2014–2018(15).Importantly,the2010agreementstatedthefollowingthreemainobjectives:toreducesickleavesothatsicknessabsencewillnotexceed5.6%,toin‐cludemorepeoplewithreducedfunctionalabilitiesintoworkinglife,andtogetpeopletoretirelater(16).Participationintheagreementisvoluntaryforthecompaniesthatmayormaynotchoosetosigntheagreementandbecomeaso‐called‘IA‐company’(7).TheIA‐agreementincludesthefollowingfivemeasures(7;16):

    1. IncreaseduseofPTSL

  • 18 Background

    2. Changesintheroleofthesickleavecertifier(e.g.,training,feedbackonownpracticesandprofessionalguidanceintheworksurroundingauthorizedsickleave)

    3. EnterprisesandbusinessescanbemembersoftheIA‐agreement4. TheFasterReturn‐to‐Work(FRW)scheme5. Changesinthefollow‐upschemesofsick‐listedemployees

    Withregardtopointfour,theFasterReturn‐to‐Work(FRW)scheme(InNorwegian:Raskeretilbake)isacollectionofmeasuresthatintendedtopreventunnecessarylong‐termsickleave(7).Theschemeimpliesoffersonindividualfollow‐up,clarificationandwork‐orientedrehabilitation;offerfortreatmentatspecialisthealthservices(purchaseofhealthservices);andlegislativeamendmentstoensurecloserfollow‐upofsickpeo‐ple.Thus,theFRWschemeisbasedontheintentionthatsick‐listedemployeesgetfasterclarification,follow‐upandwork‐orientedrehabilitationthroughNAV.TheDirec‐torateofHealthandtheLaborandWelfareDirectorateareresponsiblefortheimple‐mentationofthescheme(17).Totesttheideathatanassessmentoffunctionalabilitiescouldstrengthenthepatient’sresources,whichinturncouldfacilitateandencourageanearlyreturntotheworkplace,aNorwegianclusterrandomizedcontrolledtrialwasconducted(18).ItevaluatedtheeffectsofteachingGPsaboutstructuredfunctionalas‐sessmentstochangetheirsick‐listingpractice,especiallyinprescribingmorePTSL.ThestudyresultsshowedthattheinterventionGPsprescribedPTSLmoreoften(oddsratio[OR]1.3,p

  • 19 Background

    showedimprovementsonworkinglifeinthecountryaftertheIA‐agreement,andcon‐cludedthattheIA‐companiesfulfilledalltheirobligationswithintheagreement,andexhibitedbettercooperation(e.g.,betterassistancefromworkinglifecentersandmorefocusonclosefollow‐upofpersonsonsickleave).Nevertheless,theevaluationfoundnoeffectsonsicknessabsenceamongIA‐companies,andseveralmethodologicaldiffi‐cultieswerediscussed(22).Continuingitsinterestsinreducingsicknessabsenceinthecountry,theNorwegianMinistryofLaborarrangedameetingwithscientistsandexpertsinApril2013inordertodiscussandreviewtheevidenceregardingthedifferentmeasurestheIA‐agreementimplied(16).AttendeeshighlightedthelackofempiricalresearchontheIA‐agreementmeasures,andencouragedfurtherresearchinthearea.Afterreviewingdatafromdif‐ferentregistriesandstudiesderivedfromthem,expertsobservedthatthesicknessab‐senceratesdecreasedintheperiodswhenuseofPTSLincreased.ThedecreasewaspartiallyattributedtoanincreasedpreferenceamongGPstowardsgradedsickleave(16).Inaddition,employeesonlong‐termsickleavelistedwithGPswhooftenusedPTSLexhibitedshortersicknessabsenceandahigherprobabilityofremainingem‐ployedtwoyearslater.Someexplanatorymechanismsdiscussedbytheexpertswerehealthbenefits,reducedriskofexpulsionfromworkandaneffectofPTSLonem‐ployee’sdisciplineandattitudetowork(16).Recentregistry‐basedanalysespublishedbyNAVobservedthatuseofPTSLdoubledintheperiod2002‐2016.Halfofthesick‐nessabsencecasesgrantedin2016weregradedsickleave,withhalfofthemgradedat50%.However,PTSLofbothhigherandlowerlevelsaremorecommonlyusedovertime.AquarterofpeopleonPTSLmoveontofulltimesickleave(FTSL),andtheriskofmovingontoFTSLishighestinthefirstfewweeksofthesicknessabsenceperiod.Moreover,theresearchersfoundthatwhilethedurationofsicknessabsencehasre‐mainedfairlystableovertime,theuseofPTSLoccursearlierthanbeforeinthecourseofthesicknessabsence,withmostofthecasesgradedalreadyfromthefirstdayofab‐sence(13).TherehavebeendebatesontheeffectsoftheincreasedemphasisonPTSL.EmpiricalresearchonPTSLhasshownlargelypromisingresults,especiallyintheNordiccoun‐tries.Forexample,datafromaFinnishnationwideregistry‐basedstudyshowedbenefi‐cialeffectsofPTSLcomparedtoFTSLonreturntoworkandworkparticipation(23).PTSLhasalsobeenassociatedwithincreasedworkretentionanddecreaseduseoffulldisabilitypensioninlong‐termassessmentsinFinland(24).Similarfindingswerere‐portedinatrialamongFinnishworkerswithmusculoskeletaldisorders(MSDs)(25).InNorway,Kannandcolleagues(26),foundadeclineintheproportionofindividualsonsickleavewhentherateofPTSLincreased,aswellasshortersicknessabsencedura‐tion.DatafromanotherNorwegianregistry‐basedstudy(27)indicatedthatsick‐listedemployeeswhowereonPTSLwhentheycompletedawork‐relatedrehabilitationpro‐gramweremorelikelytoreturntoworkcomparedtothosewhowereonFTSLwhentheylefttherehabilitationclinic.InSweden,PTSLwasfoundtobeassociatedwithanearlierreturntoworkinpeoplewithmentaldisorders(MDs)after60daysofFTSL(28),whilstnoeffectofPTSLonearlyreturntoworkwasobservedamongpeoplewithMDsinanotherstudyconductedinDenmark(29).Grasdal(30),whoreviewedahand‐

  • 20 Background

    fulofempiricalstudiespublishedpriorto2016,concludedthatoverall,theresultsindi‐catedthatPTSL“contributestoreductioninsicknessabsence.Specifically,gradingseemstocontributetoreducingthesickleaveperiod,butthereisconsiderableuncer‐taintyaboutthesizeofthiseffect”(p.114).Althoughnotablepromisingresultshavebeenpublished,researchontheeffectsofPTSLhasbeencriticizedforhavingweakexternalvalidity,andmethodologicalflaws,suchasselectionbias,highuseofself‐reporteddata,andweaknessesinmeasuringworkparticipation(1;31).Todate,nosystematicanalysisoftheevidencebaseontheeffectsofPTSLcomparedtoFTSLhasbeenundertaken.Therefore,thissystematicmappingreviewaimedtomapallquantitativeevidenceontheeffectsofPTSLversusFTSLonsicknessabsenceandworkparticipation.

  • 21 Methods

    Methods

    Weconductedasystematicmappingreviewtoanswerthequestion:whatevidenceex‐istsandwhatdoesitsayabouttheeffectsofPTSLversusFTSLonsicknessabsenceandworkparticipation?Theprojectteam(reviewers)andcommissioner(NAV)discussedandagreedontheresearchprotocol,whichisavailableuponrequest.

    Whatisasystematicmappingreview?

    Systematicmappingreviews(alsoknownassystematicscopingreviews)arereviewsthatmapanddescribetheexistingliteratureorevidencebaseonaparticulartopic(32).Suchliteraturereviewstakestockoftheresearchavailableinaparticularfield.Thistypeofreviewproducesausefulendproductinitsownright,describingtheem‐piricalresearchthathasbeenundertakenwithinaparticularfieldofstudy,butalsoprovidesanoverviewofaresearcharea,highlightingwhereempiricalresearchislo‐catedandwheretherearegaps.Itdoesnotincludeasynthesisofindividualstudyre‐sults(32;33).Inatypologyofreviews,GrantandBooth(34)explainthatsuchreviews“mapoutandcategorizetheexistingresearchonaparticulartopic,identifyingresearchgapsfromwhichtocommissionfurtherreviewsand/orprimarystudies.”ThepresentmappingreviewwasmethodologicallyguidedbyaframeworkproposedbyArkseyandO’Malley(32),aswellasLevacandcolleagues’(35)recommendationsonclarifyingandenhancingeachstageofthereview.Thus,themethodologicalstepswere:

    1.Identifyingtheresearchquestion2.Identifyingrelevantstudies3.Selectingstudies4.Chartingthedata5.Collating,summarizingandreportingtheresults6.Optionalconsultation.

    Asseenfromthisoutlineofthemethodologicalstepsofsystematicmappingreviews,qualityappraisalisnotadefinedstepwithinsuchreviews(32;33).However,wede‐cidedtoappraisethemethodologicalqualityoftheincludedstudiesuponagreementwiththecommissioner(NAV).Qualityappraisalwasperformedaspartofstep4,chart‐ingthedata.ThissystematicmappingreviewisreportedinaccordancewiththePRISMA‐ScRreportingguideline(36).

  • 22 Methods

    Selectioncriteria(identifyingtheresearchquestion)

    Theselectioncriteriawerediscussedandagreedwiththecommissioneraheadoftheliteraturesearch.WeincludedquantitativeresearchaddressingtheeffectsofPTSLver‐susFTSLonsicknessabsenceandworkparticipation.Themainstudyinclusioncrite‐rionwasasubstantialemphasisontheeffectofPTSLversusFTSLasthesubjectmatter.TheselectioncriteriawereguidedbythefollowingPICO(population,intervention,comparison,andoutcome)elements:Population:Part‐timeorfull‐timeadultemployees(16‐69yearsold).Weexcludedstudiesofpeoplewhoweredescribedasself‐employed.Intervention:Partialsickleave(PTSL).Followingthe2015reportbytheNordicSocialStatisticalCommittee(7),wedefinedPTSLasgradedleave,thatallowspeoplewithre‐ducedworkabilitytoworkparttime.PTSLvariesbetween20%upto99%(7;12).WepresentanydifferencesintheconceptualizationofPTSLacrosstheincludedstudiesintheresults.Comparison:Full‐timesickleave(FTSL),i.e.,nophysicalpresenceattheworkplace.Outcomes:

    Sicknessabsence(extent,duration,andsimilar):man‐dayslostduetoownsick‐nessasapercentageofcontractualman‐days(6).Oneman‐daycorrespondstothelengthintimeofoneworkingdayforapersoninafull‐timeposition(100%position).Workparticipation(extent,beingfired,andsimilar)measuredaspositionpro‐portion(valuesmightrangebetween0to1)(6).Degreeofdisabilityandrehabilitationbenefits:whetherthepersoninques‐tionisonfull‐orPTSL,andtowhatdegree,isindicatedbythedegreeofdisability.Thedegreeofdisabilityisbetween20%and100%,where100%meansfulldisa‐bility.Ifasicknessabsencecaseconsistsofmorethanonemedicalcertificate,re‐searchersmightreporttheaverageofthedegreesofdisabilityforthemedicalcertificatesinquestion(6).Health‐relatedoutcomes:Diseaseseverityanddisability.

    Studydesign:Duetothefactthatthiscommissionwastiedtoaneffectivenessques‐tion,weaimedtoincluderandomizedcontrolledtrials(RCTs)aswellasnon‐random‐izedstudieswithacontrolcondition.Weincludednon‐randomizedstudiesbecauseweanticipatedthatfew,ifany,RCTshadbeenconductedinthisfield.Registry‐basedstudies(RBs)(alsoknownaspaneldataanalysis)wereincludedinthisreviewafterroundsofconsultationwiththecommissioner.RBsinvolvethestatisticalanalysisofdatasetsfromregistriescontainingmultipleobservationsovertimeofasamplingunit(37).RBscanbeconductedbypoolingtime‐seriesobservationsacrossavarietyofcross‐sectionalunits,includingindividuals,countries,orcompanies(37;38).

  • 23 Methods

    However,thesestudiesdonotenableresearcherstoestablishcausalrelationshipsamonganinterventionorexposureandoutcomes(38).Insum,weconsideredthefollowingstudydesignsforinclusion:

    Randomizedcontrolledtrials(RCT) Non‐randomizedcontrolledstudies/Quasi‐randomizedstudies(NRCT) ControlledBefore‐Afterstudies(CBA) InterruptedTimeSeries(ITS)withatleastthreemeasurementpointsbefore

    andaftertheintervention Registry‐basedstudies(RBs)

    Wefollowedthedefinitionsofnon‐randomizedstudiesproposedbytheEffectivePrac‐ticeandOrganisationofCareCochraneGroup(39)(seeGlossaryinAppendix1).Publicationdate:Studiespublishedbetween1990and2018.Language:WeincludedalllanguagesaslongastherewasanabstractinEnglish.Anystudiesmeetingtheinclusioncriteriaandpublishedinlanguagesnotmasteredbythereviewteam(English,Spanish,Norwegian,Swedish,Danish,German)wouldhavebeentranslatedwithGoogletranslateorbyacolleagueattheNIPH.Context:Studiesconductedinhigh‐incomeeconomies(GNIpercapitaof$12,236ormore)asdefinedbytheWorldBank(Link:https://data‐helpdesk.worldbank.org/knowledgebase/articles/906519‐world‐bank‐country‐and‐lending‐groups).Literaturesearch(identifyingrelevantstudies)

    Afterextensivedialoguewiththecommissionertoagreeontheresearchquestionandtheselectioncriteria,aresearchlibrarian(ElisabetHafstad)plannedandexecutedsys‐tematicsearchesinthefollowingdatabases(frominceptiontoJanuary2018):

    CochraneLibrary:CENTRAL Embase MEDLINE PsycINFO PubMed SociologicalAbstracts&SocialSciencesAbstracts SveMed+ WebofScience

    Thesearchstrategywasadaptedforeachdatabase.Thefinalsearchstrategyispro‐videdinAppendix2.SearchingothersourcesToidentifyadditionalstudies,wehandsearchedthebibliographiesofallincludedstud‐ies,aswellasanyliteraturereviewsandseminalreportsaboutPTSL.Wesearchedthe

  • 24 Methods

    websiteoftheNorwegianInstituteofPublicHealth,Idunn(NordicJournalsonline),theNorwegianandNordicindextoperiodicalarticles(Norart),OpenGrey,Google,andGoogleScholarandscreenedthefirst200hits.Tworeviewers(JM,RB)alsohandsearchedontheNordicLabourJournal(http://www.nordiclabourjournal.org/),theCampbellLibraryandthefollowingwebsites:Nordicorganizationsforlaborandworkenvironment

    • TheDanishNationalResearchCentrefortheWorkingenvironment• TheDanishAgencyforLaborMarketandRecruitment• FinnishInstituteofOccupationalHealth• TheNorwegianLaborandWelfareAdministration(NAV)• TheNorwegianMinistryofLaborandSocialAffairs• Försäkringskassan(Sweden)• TheSwedishMinistryofHealthandSocialAffairs

    Ongoingandrecentlycompletedclinicaltrials

    WorldHealthOrganizationInternationalClinicalTrialsRegistryPlatform(http://www.who.int/trialsearch/)

    NationalInstituteofHealthclinicaltrialsdatabase(http://clinicaltrials.gov)Studyselection

    Allrecordsretrievedthroughtheliteraturesearcheswereindependentlyscreenedforeligibilityagainsttheselectioncriteriabytworesearchers(JMandNB)byusingapre‐designedscreeningform.Wefirstscreenedtitlesandabstractsandthenproceededtofull‐textscreeningofrelevantrecordstodecidefinalinclusionorexclusion.Inclusionwasdecidedbyconsensusandanydiscrepanciesweresolvedbydiscussion.Ifneces‐sary,wewouldhaveinvolvedathirdresearcher(RB)tosolvediscrepancies.Dataextraction(chartingthedata)

    Wedesignedadataabstractionformtogatherrelevantinformationfromeachstudy,includingcharacteristicsofstudyparticipants,settings,context,percentageofsickleaveevaluatedinthestudy,comparisons,studydesigns,methods,statisticalanalysesandcovariates,andresults.Onereviewer(JM)extractedalldatafromtheincludedstudiesandasecondreviewer(NB)checkedtheinformationforaccuracyandcom‐pleteness.HNandDTKassistedwithcheckingdataaccuracy.Disagreementsweresolvedbydiscussion,consensus,andparticipationofRB.Qualityappraisaloftheincludedstudies

    Tworeviewauthors(JM,NB/RB)appraisedthemethodologicalqualityofeachin‐cludedstudyindependently.Weresolveddisagreementsbyconsensus.Ifnecessary,wewouldhaveinvolvedanotherresearcher.

  • 25 Methods

    RCTswereappraisedbyusingtheCochranetoolforassessmentofriskofbiasofRCTs(40).Thus,weassessedthefollowingcriteria:

    •Randomsequencegeneration(selectionbias).•Allocationconcealment(selectionbias).•Blinding(performancebiasanddetectionbias),blindingofparticipantsandpersonnelassessedseparatelyfromblindingofoutcomeassessment.•Incompleteoutcomedata(attritionbias).•Selectivereporting(reportingbias).•Otherbias.

    WecriticallyappraisedtheRBsbyusingthechecklistforcohortstudiesdescribedintheNIPHhandbook‘Slikoppsummererviforskning’(41).This10‐itemschecklistevalu‐atesknownsourcesofbias,suchasselectionbias,incompleteorlackofreportingofoutcomeassessment,dropouts,confoundingfactors,andblindingofoutcomeassess‐ment.WeusedthistoolbecauseitisthemostsuitableappraisaltoolwecouldidentifyforRBs.Wesearchextensivelyandaskedmethodologicalexperts,andtheirrecommen‐dationwastousethechecklistforcohortstudies.IthasbeenusedbyustoappraiseRBsinprevioussystematicreviews.Whilethischecklisthaslimitations,tothebestofourknowledge,thereisnouniquechecklistforappraisingRBs.Intheeventthatnon‐randomizedcontrolledstudies,includingCBAsandITSs,hadbeenincluded,wewouldhaveappraisedthemethodologicalqualityofsuchstudieswiththetoolsuggestedbytheCochraneEPOCGroup(42).Thistoolincludesadditionalitems(relativetotheCochraneriskofbiastoolforRCTs)toassesstheriskofselectionbiasandsubsequentconfounding.Theadditionalitemsare“werebaselineoutcomemeas‐urementssimilar?”and“werebaselinecharacteristicssimilar?”(42).Collatingandsummarizingtheresults

    Asdescribedabove,mappingreviewsprovideanoverview‐anddescriptionofexistingresearch.Datasynthesisislimited,relativetofullsystematicreviews:Asystematicmappingreviewdoesnotincludeasynthesis,suchasmeta‐analysis,ofindividualstudyresults.Inaccordancewiththeaim‐andmethodologicalscopeofsystematicmappingreviews,weanalysedthedatadescriptively,withfrequenciesandpercentages,andpresentedresultsintext,tables,andfigures.Wegroupedstudiesintocategoriesac‐cordingtohowtheywereseentorelatetoeachother,followingadatadrivenap‐proach.Forclarity,wepresentedinformationseparatelyforRCTsandRBsaswellasresearchfromNordiccountries.WenotethatRBsdonotenableresearcherstoestab‐lishcausalrelationshipsamonganinterventionorexposureandoutcomes.

  • 26 Results

    Results

    Searchresults

    Theelectronicsearchesinthemajordatabasesyielded676references,andadditionalsearchesingreyliteraturesourcesadded30references.Atotalof300duplicateswereremoved.Weexcluded380outofthe406referencesscreenedattitle/abstractlevel,andwereadtheremaining26referencesinfull‐text.Thirteenstudiesmetourinclusioncriteria.Threeoutofthe13includedstudieswereidentifiedafterconsultinglaboragenciesandinternationalministries(15;43;44).Figure1depictstheflowdiagramfortheselectionofthestudies.Norelevantongoingstudieswereidentifiedbysearchinginthetrialregistries.Thein‐terventionsunderevaluationintheregisteredprotocolsaboutreturn‐to‐worktrialsin‐cludedbehavioralinterventions,self‐management,psychotherapy,motivationalinter‐viewing,andotherintegratedactivereturn‐to‐workprograms.Excludedstudies

    Mostofthe13excludedstudiesreadinfull‐texteitherdidnotevaluatetheeffectsofPTSLorwerenotempiricalresearch.ANorwegianRCTthatevaluatedactivesickleavedidnotmeettheinclusioncriteriaastheauthorsexcludedemployeesonpartialsickleave(45).Thistrialfoundthatactivereturn‐to‐workimprovedneitherthenumberofdaysonsickleavenortheproportionofpatientsreturningtoworkinworkerswithlowbackpainfrom65Norwegianmunicipalities(45).Additionally,twoNorwegianRBswereexcludedbecauseofthelackofacomparisongroupwhowereonFTSL.Thesetwostudiesevaluatedsolelydatafromsick‐listedemployeeswhoreceivedPTSL(26;46).Weprovidethemainreasonsforexclusionofthe13referencesexcludedafterfull‐textconsiderationinAppendix3.

  • 27 Results

    Figure1.Flowdiagramoftheselectionofstudies Descriptionofincludedstudies

    Theevidencepresentedinthissystematicmappingreviewconsistsof13studies.WeincludedoneRCTfromFinland,reportedintwopublications(25;47),whiletheremain‐ing12studieswereRBs(1;15;17;24;28;43;44;48‐52).Thesestudiesusedobserva‐tionalmethodstoexploretheinteractionsbetweenPTSLanddifferentvariables,suchasreturn‐to‐work(RTW),insetsofpaneldatafromregistries.ResearchaimOverall,all13includedstudiesexaminedtheeffectsofPTSLcomparedtoFTSLforsick‐listedemployees(table2).

    References screened at title and abstract (n = 406) 

    References identified from the  database searches 

    (n = 676) 

    Additional references identified from other sources  

    (n = 30) 

    References after duplicate removal (n = 406) 

    References excluded (n = 380) 

    References screened in full‐text (n = 26) 

    References excluded  (n = 13) 

    Included studies  (n = 13) 1 RCT 12 RBs 

  • 28 Results

    RandomizedcontrolledtrialResearchersoftheFinnishInstituteofOccupationalHealthconductedtheonlyRCTin‐cludedinthismappingreview.Thetrial,whichfollowedaprospectiveparalleldesignandwasreportedintwodifferentpublications,evaluatedtheeffectsofearlyPTSLonRTWandsicknessabsence(25),andonhealth‐relatedoutcomes(47)amongworkerswithMSDs.Registry‐basedstudiesAllthe12includedRBsevaluatedtheeffectsofPTSLcomparedtoFTSLforsick‐listedemployees.WenotethattwoGermanstudiesevaluatedemployeeswhocompletedarehabilitationprogram,whichismandatoryinthecountry(17;50).Twootherstudies,fromFinland,addressedthetransitiontodisabilitypension(24)andtheintroductionofnewlegislationofPTSL(1).Lastly,oneRB(24)hasacompanionpaperwithanalysesofthesamedataset,whichweappliedwhenrelevant(23).Setting

    Ingeneral,itcanbestatedthatmostoftheevidencebaseontheeffectsofPTSLcomesfromNordiccountries,as11studies(85%)weredoneinsuchsettings(1;15;24;25;28;43;44;47‐49;51).ThetworemainingRBswereconductedinGermany(17;50).Seetable2. Table2.Countryandresearchaimoftheincludedstudies(n=13)Study,year Country ResearchaimAndrén2012(48)

    Sweden ToexaminethebenefitsofbeingonPTSLcomparedtoFTSLinindividualswithmusculoskeletaldisorders.

    Andrén2014(28)

    Sweden ToanalyzetheimpactofPTSLontheprobabilityofreturningtoworkwithfullrecoveryoflostworkcapacitywithin1yearforemployeeswithmentaldisorders.

    Bethge2016(17)

    Germany TodeterminetheeffectsofPTSLondisabilitypensionandreg‐ularemploymentinarandomsampleofrehabilitationpatientswhofinishedarehabilitationprogrambetween2002and2009.

    Grødem2015(15)

    Norway Tostudyemployeeswhocompletedtheirperiodofsicknessbenefits(2‐3yearsaftercompletion)butarestillunabletofullyreturntowork.

    Høgelund2010(52)

    Denmark Toexaminetheeffectofanationalgradedreturn‐to‐workpro‐gramontheprobabilityofsick‐listedworkersreturningtoreg‐ularworkinghours.

    Kausto2012(24)

    Finland ToestimatetheeffectsofPTSLonthe transitiontodisabilitypensionapplyingpropensityscoremethods.

    Kausto2014(1) Finland ToexaminetheeffectsofthenewlegislationonPTSLonworkparticipationofemployeeswithlong‐termsicknessabsence.

    Lie2014(43) Norway ToevaluatetheeffectsofPTSLvsFTSLonsicknessabsence.Markussen2012(49)

    Norway ToexaminewhetherPTSLcanreduceabsenteeismandsubse‐quentsocialinsurancedependency,andpromoteself‐suffi‐ciency.

    Nossen2013(44)

    Norway ToexploretheroleofdifferentdefinitionsofPTSLcomparedtoFTSLonsicknessabsenceduration.

  • 29 Results

    Shiri2013(47)Viikari‐Juntura2012(25)RCT

    Finland Todeterminethehealth‐relatedeffectsofearlyPTSLamongemployeeswithmusculoskeletaldisorders(47).ToevaluatetheeffectsofearlyPTSLonreturntoworkandsicknessabsenceamongpatientswithmusculoskeletaldisor‐ders(25).

    Streibelt2017(50)

    Germany TodeterminetheeffectofPTSLinadditiontoamultimodalre‐habilitationprogramonlong‐termworkparticipationinpeoplewithchronicmentaldisorders.

    Viikari‐Juntura2017(51)

    Finland ToassesstheeffectivenessoftheuseofPTSLattheearlystageofworkdisability(first12weeks)duetomentaldisorderormusculoskeletaldiseaseonsustainedreturntowork(RTW)andoverallworkparticipation.

    Typeofpublicationandpublicationyear

    Mostofthestudiesincludedinthissystematicmappingreviewwerepublishedinpeer‐reviewedjournals(11studies,85%).TwoNorwegianRBswerepublishedasorganiza‐tionalreports(15;43).TheFinnishRCTwaspublishedin2012‐2013(25;47).TheRBswerepublishedbetween2010and2017,withmostpublishedaround2014.ParticipantsBelow,wepresentthecharacteristicsoftheparticipantsintheRCTandRBsseparately.RandomizedcontrolledtrialTheFinnishtrial(25;47)included62employeesonsickleaveduetoMSDs.Theywererecruitedfromsixoccupationalhealthunitsofmedium‐andlarge‐sizeprivateorpublicenterprises.Theyhadapermanentorlong‐termcontract,workingfull‐time(37‐38hoursperweek),andhadnotbeenonsickleaveduetotheirMSDsfor>2weeksduringtheprecedingmonthand>30daysduringthepreceding3months.Mostofthepartici‐pantsworkedinthehealthcaresectororretailtrade,andaminorityfromcall‐centresormeat‐processingindustry.Aroundhalfoftheparticipantshadhighervocationalschoolbutnoneofthemhadcompleteduniversitystudies(table3).Table3.Characteristicsofparticipants,interventionandcontrolgroupintheFinnishrandomizedcontrolledtrialRandomizedcon‐trolledtrial(RCT)

    Population Intervention:Part‐timesickleave,

    PTSLN(%)

    Comparison:Full‐timesickleave,FTSLN(%)

    Viikari‐Juntura2012(25),Shiri2013(47)Finland

    N=62employeesAge:meanage44(standarddeviation10)Sex:97%femaleDiagnosis:Musculoskeletaldisorders(e.g.,neckorshoul‐derregion,backorupperorlowerextremities)Ethnicity:notreported

    N=31(50%)70%received50%PTSL30%workedshorterhourson3–4daysaweek

    N=31(50%)

  • 30 Results

    Registry‐basedstudiesTheparticipantsinthetwelveregistrystudieswerefromNorway(fourstudies),Den‐mark(onestudy),Finland(threestudies),Sweden(twostudies),andGermany(twostudies).Intotal,2,741,563participantswereanalyzedintheseRBs(range627–1,400,094).Ingeneralterms,thevastmajorityofparticipantsacrosstheRBssharedcommoncharacteristics,suchashavingaregularjobcontract,beingfemale(upto78%)andaround45yearsold(range18‐64).GrødemandcolleaguesincludedNorwe‐gianemployeesolderthan60(15),whileLie(43)andNossenandBrage(44)didnotprovideinformationonparticipants’age,genderordiagnosis.OnlythetwoSwedishstudiesprovideddataonparticipants’ethnicity(28;48).Seetable4.NationalregistriesusedforanalysisAlltwelveRBsuseddatafromnationalregistries.ThefourstudiesconductedinNor‐wayallutilizeddatafromtheNorwegianLaborandWelfareAdministration(NAV),asfollows:

    Markussenanalyzeddatafromallsicknessabsencesgrantedfrom2001through2005(49).

    NossenandBrageanalyzeddatafromallspellsgrantedin2011(44). Lie(43)analyzeddatafromspellsgrantedto10%ofallpeoplereportedinNAV’sregistryduring2002‐2010.

    Grødem(15)analyzeddatafromemployeeswhoterminatedthesicknessbenefitsperiodduringthefirsthalfof2011afteroneyearofsickleave.

    ThissuggestssomeoverlapinthecasesbasedontheNAVregistry,betweenLie(43)andMarkussenandcolleagues(49)andbetweenNossenandBrage(44)andGrødemandcolleagues(15).Weprovidefurtherdescriptionsinthesection“partialsickleaveintheincludedstudies”below.Therewerenoindicationsofoverlapbetweendatasourcesintheremainingstudies.Andrénandcollaboratorsusedthesamedatasetfromthe2002sampleoftheSwedishSocialInsuranceAgencycontainingdataofsick‐listedemployeesduetoMSDs(48)andmentaldisorders(MDs)(28).Høgelundandcol‐leaguesuseddatafromthenationalregisterofpaymentsofsicknessbenefitsinDen‐mark,supplementedwithsurveyinformation(52).InthetwoGermanstudies,Bethgeandcolleagues(17)studiedemployeeswhocontinuedonsickleaveaftertheycom‐pletedarehabilitationprogram(January‐June2007),whileStreibeltandcolleagues(50)in2012recruitedemployeeswhohadcompletedarehabilitationprogramandwereeligibleforPTSL.BothstudiesuseddatafromtheGermanPensionInsuranceAgency.Table4.Characteristicsofparticipants,interventionandcontrolgroupsinregistry‐basedstudies(n=12)Registry‐basedstudies

    Population Part‐timesickleave,PTSLN(%)

    Full‐timesickleave,FTSLN(%)

    Andrén2012(48)

    N=1170;SwedenAge:20‐64years(50%>46years)Diagnosis:MSDsGender:around60%femaleEthnicity:85%borninSweden

    N=140(12%) N=1030(88%)

  • 31 Results

    Andrén2014(28)

    N=627;SwedenAge:20‐64years(50%>46years)Diagnosis:MDsGender:around60%femaleEthnicity:85%borninSweden

    N=79(13%) N=548(87%)

    Bethge2016(17)

    N=3,750;GermanyAge:average45years(range18‐60)Diagnosis:Around63%MSDs.Otherdi‐agnoses:cardiac,oncological,psychoso‐maticGender:around50%femaleEthnicity:notreported

    N=1875(50%) N=1875(50%)

    Grødem2015(15)

    N=17,077;NorwayAge:>60years(bornafter1951)Diagnosis:MSDs(45.5%),MDs(25.4%)Gender:58%femaleEthnicity:notreported

    N=5294(31%)>50%PTSL(8.7%)

    50%PTSL(14.6%)

    50%PTSL,4.2%50%PTSL,2%

  • 32 Results

    Streibelt2017(50)

    N=762;GermanyAge:average47.8yearsDiagnosis:MDs(65%affectivedisorders)Gender:78%femaleEthnicity:notreported

    N=381(50%) N=381(50%)

    Viikari‐Juntura2017(51)

    N=3756;FinlandAge:20‐64(41%20‐44years,37%45‐54years,22%55‐64years)Diagnoses:MDsandMSDsGender:77.5%femalesEthnicity:notreported

    N=1878(50%) N=1878(50%)

    MSD=Musculoskeletaldisorder,MD=Mentaldisorder,SD=standarddeviationIntervention(partialsickleave=PTSL)

    Inthesectionbelow,wepresentthecharacteristicsoftheintervention,PTSL,intheRCTandRBsseparately.RandomizedcontrolledtrialIntheFinnishRCT(25;47),theGPsgavethepatientafitnote,indicatingthedurationofpartialworkdisability,whethercertainphysicalloadsshouldbereduced,andwhetheranyadditionalworkmodificationsweredeemednecessary.A50%PTSLwasgivento70%ofallsick‐listedemployees,whereasduetodifficultiesinarrangingahalfworkday,30%ofemployeesworkedshorterhourson3–4daysaweek.Sometaskmodifica‐tionswerealsoimplementedifnecessary.Registry‐basedstudiesInthe12RBs,alldataontheuseofPTSLweretakenfromthenationalregistriesintherespectivecountries(seeaboveNationalregistriesusedforanalysis).Themostcom‐monlyusedPTSLwas50%,whichwasgrantedinaround70%ofthegradedworkab‐sencesformostofthetime.ThestudyfromDenmarkdescribesPTSLasaworkplacein‐terventionprogramwhereby“sick‐listedworkersreturntotheirpre‐sickleavejobontemporarilyreducedworkinghours”(52).TheRBsanalyzeddataaboutsicknessab‐sences(bothPTSLandFTSL)grantedbetween2001and2014,whichgivesatimelineof13years.Figure2illustratesthedispersionoftheyearsfordataanalysisacrosstheRBs(i.e.,thetimelinefromwhichthedataweretakenintheregistry).ThetwoSwedishRBs,bothwithAndrénasfirstauthor,usedthesameyearsfromtheregistry,butin‐cludedworkerswithdifferentdisorders,MSDsandMDs,respectively(28;48).ItisimportanttohighlightthatallRBs,exceptfortwoNorwegianstudiespublishedbyGrødemandcolleagues(15)andNossenandBrage(44),exploredtheinfluenceofdif‐ferentcovariatesontherelationshipbetweenPTSLandtheoutcomemeasures.MostoftheRBsconductedpropensity‐scorematchinganalysesforbalancingthesamplesontheprobabilityofbeingassignedtoPTSL(sixstudies).ThemostcommoncovariatesusedacrosstheRBsweregender/sexandage(fivestudies),followedbytypeofoccu‐pationanddiagnosis(fourstudies),andgeographicarea,incomeanddataonthephysi‐cianwhograntedthesickleave(threestudies).Othercovariatesincludedthesickleaveduration,previoussickleaves,levelofeducation,etc.DetailsontheadjustedanalysesandthecorrespondingcovariatesacrossRBscanbefoundinAppendix4,anddefini‐tionsofthestatisticalanalysesarepresentedinAppendix5.

  • 33 Results

    Figure2.Dispersionoftheyearsfordataanalysisacrosstheregistry‐basedstudies(n=12)Comparison

    BoththeFinnishRCTandthe12RBsusedFTSLascomparator. Outcomemeasures

    Intotal,15outcomeswerereportedinthe13includedstudies.Therewerefourmaintypesofoutcomes:sicknessabsence,workparticipation,degreeofdisability,andhealth‐relatedoutcomes.Workparticipation(RTW)wasthemostcommonoutcome,reportedintenstudies(i.e.,intheRCTandnineRBs),followedbysicknessabsencedu‐ration,disability,disabilitypensionandallowanceofsocialbenefits(eachmeasuredbyfourstudies).Because12ofthe13includedstudieswereRBs,registrydatawerethemostcommonsourceofmeasurement.Table5showsthedifferentoutcomes,numbersofstudiesandtools.Table5.OutcomemeasuresandtoolsintheincludedstudiesOutcomemeasure Numberofstudies ToolsSicknessabsenceduration

    1RCT(25)3RBs(44;49;50)

    Registrydata

    WorkparticipationReturn‐to‐work(RTW)

    1RCT(25) Registrydata

  • 34 Results

    9RBs(1;15;28;43;48‐52)

    Unemployment 3RBs(49‐51) RegistrydataDegreeofdisabilityRecurrenceofsickleaveforanycause

    1RCT(25)

    Registrydata

    Disability 1RCT(47)3RBs(15;43;51)

    RegistrydataShirietal.2013(47)usedthefollowingtools:OswestryDisa‐bilityIndextoassessthedisabilitylevelduetobackpain;theNeckDisabilityIndextoassesscervicalspine‐relateddisabili‐ties;theQuickDASHtoassessthedisabilitiesofthearm,shoulder,andhand;andtheComprehensiveOsteoarthritisTest(COAT)toassessthesymptomsofthehiporknee.

    Productivityloss 1RCT(47) TwoquestionsrecommendedbyBrouweretal.,“Thesubjectswereaskedtoconsiderthelatestfullorpartialworkingdayandcompareittotheirnormalworkdaywhenansweringthequestions:(i)assesstheamountofworkyouwereabletoper‐form,and(ii)assessthequalityofyourwork.Forbothques‐tions,thescalerangedfrom0–10(0=verypoorto10=regularquantityorquality).Incasethereportedvaluewas

  • 35 Results

    Qualityappraisaloftheincludedstudies

    Whilenotarequiredstepinasystematicmappingreview,weassessedthemethodo‐logicalqualityofthe13includedstudies.WeuseddifferentchecklistsfortheRCTandthe12RBs(seemethods).RandomizedcontrolledtrialTheFinnishRCTexhibitedmoderatemethodologicalquality(25;47).Thereweresomeconcernsaboutbothperformanceanddetectionbiasduetothelackofblindingofbothparticipantsandoutcomeassessor.Inaddition,thetrialreportedsomeoutcomesthatwerenotpre‐specifiedintheprotocol.Wefoundnomajorconcernsforselectionorat‐tritionbias.SeeAppendix6foracompletedescriptionofthequalityappraisaloftheRCT.Registry‐basedstudiesWeusedachecklistfromtheNIPHhandbookforsystematicreviews(41)toassesstheoverallqualityoftheRBs,whichresultedinacategorizationofstudiesintolow,moder‐ateorhighmethodologicalquality,asfollows:

    Highquality:lowriskofbiasin≥8items. Moderatequality:lowriskofbiasin5‐7items. Lowquality:lowriskofbiasin≤4items.

    ElevenoutofthetwelveincludedRBsexhibitedmoderatemethodologicalquality(1;15;17;24;28;43;44;48;49;51;52).Oneshowedhighmethodologicalquality(50).Allstudiesincludedlargenumbersofindividuals,werejudgedasrepresentativeoftheirpopulation,andmeasuredbothexposureandoutcomesequallyandreliably.ThePTSLandFTSLgroupswerecomparableonimportantbackgroundfactorsinfivestudies(1;15;28;50;51).Follow‐uptimewasjudgedasadequateinallstudies.SeeAppendix7foracompletedescriptionofthemethodologicalappraisaloftheRBs.Summaryofmainfindingsfromtheincludedstudies

    Weprovideabriefsummaryofthemainfindingsreportedbythestudyauthorsfortheoutcomesincludedinthisreview,i.e.,sicknessabsence,workparticipation,degreeofdisabilityandrehabilitationbenefits,andhealth‐relatedoutcomes.Furtherdatacanberetrievedinthefull‐textpublications.Asummarytable,indicatingdirectionofresultsacrossthestudies,isprovidedattheendofthechapter(table6).Sicknessabsenceduration

    TheFinnishRCTofmoderatemethodologicalquality,andthreeRBs,ofmoderateandhighmethodologicalquality,reportedonsicknessabsence.TheFinnishtrialshowedalowerproportionofsicknessabsencedaysinemployeeswithPTSLthaninpeerswithFTSLthroughoutthe12‐monthfollow‐upperiod(20%loweronaverage)(25).

  • 36 Results

    ThreeRBsmeasuredsicknessabsenceduration.TheNorwegianRBpublishedbyMarkussenandcolleagues(49)foundthatPTSLpredictedreductionsinsicknessab‐sencedurationbymorethan60daysaftercontrollingforpatient/jobcovariatesandphysiciancharacteristics,comparedtoFTSL.Thisreductionincreasedupto74dayswhenspellsexceeding12weekswereanalyzed.AnotherNorwegianstudy,byNossenandBrage(44),observedthattheuseofPTSLinspellslastingatleast15daysandthatwereonFTSLduringthefirst14daysledtoa21daysshortersicknessabsencedura‐tioncomparedtotheuseofFTSL.LargerreductionsinfavorofPTSLwereseenafter8weeksinspellslastinglongerthan2weeksandgradedduringthefirst2weeks(166daysinPTSLvs199daysinFTSL;meandifference=‐33days).ThisdifferenceinfavorofPTSLincreasedupto39daysat12weeks.Conversely,crudeanalysesrevealedshortersicknessabsencesamongthoseassignedtoFTSLcomparedtothoseinPTSL(115daysinPTSLvs23daysinFTSL;meandifference=92days).Thisdifferencewasreducedto13dayswhendataforspellsgradedafter2weekswereanalyzed1.Theanal‐ysisofspellsgradedat

  • 37 Results

    morethanoneandahalftimesaslikelytoRTWthanthoseonFTSL.Controllingforprevioussicknessabsenceduringthepreceding30daysdecreasedtheHRby11%andcontrollingforbodymassindexdecreaseditby3%.OverallHRforRTW,controllingforage,paininterferencewithsleep,andprevioussicknessabsence,was1.76(95%CI1.21–2.56).ThreeoftheNorwegianRBsreportedonworkparticipation.Grødemandcolleagues(15)reportedthatemployeesonPTSLstayedconnectedtotheirjobsmorethantheirpeersonFTSLattwoyearsafterterminationofthesicknessbenefits(38%inFTSLvs84%in

  • 38 Results

    sociatedwithagreaterlikelihoodoffullrecoverycomparedtoFTSL(averagetreat‐menteffect0.015)2whenitisassignedinthebeginningofthespell.Andrénalsoob‐servedastrongpositiveeffectofPTSL(averagetreatmenteffect0.387)2,andstatisti‐callysignificant,whenassignedafter60daysofFTSLattheendofthe330daysobser‐vationperiod.DistributionalanalysisoftheeffectparametersshowedthatinagroupofrandomlyselectedemployeesonsickleaveformorethantwoweeksduetoaMD,17.8%ofthemwouldfullyrecovertheirlostworkcapacityifassignedtoPTSLinthebeginningofthespell,butwouldnothavefullyrecoveredtheirlostworkcapacitywith‐outthePTSLtreatment.However,16.3%ofthemwouldnotfullyrecoverifassignedtoPTSL.TheotherthreeRBsthatreportedonworkparticipationwerefromFinland(twostud‐ies)andGermany.InFinland,Kaustoandcolleagues(1)observedreductionsinthelevelofworkparticipationforboththePTSLandtheFTSLgroupsduringtheone‐yearfollow‐up,theabsolutereductionbeinglargerintheFTSLgroup(−26.5%)ascomparedwiththePTSLgroup(−21.2%),whichmeansadifferenceof5.3%(95%CI3.1%to7.5%).Thisdifferenceincreasedupto9.8%(95%CI5.9%to13.7%)inthepropensityscorematchedsubsample(i.e.,theconditionalprobabilityofbeingassignedtoPTSLgivenobservedcovariates).Subgroupanalysesshowedthatinallagecategories,workparticipationdeclinedmoreintheFTSLgroupthaninthePTSLgroup.Thedifferenceinthedeclinewassignificantinagecategories45–54and55–65.Therewasnoeffectinthoseaged35–44.Insubgroupanalyses,astatisticallysignificantlylargereffectinfa‐vorofPTSLwasfoundforpeoplewithmentaldisordersascomparedwiththeotherdi‐agnosticcategories(difference12.8%,95%CI9.0%to16.5%).AnotherFinnishstudypublishedin2017byViikari‐Junturaandcolleagues(51)foundanabsoluteriskdifferenceof8.0%andarelativeriskdifferenceof10.9%infavorofPTSLonsustainedRTW.Inaddition,theauthorsobservedthatthemeanoveralltimespentatworkwas77.4%;itwas10.5%higherinthePTSLgroupcomparedtotheFTSLgroupduringthe2‐yearfollow‐up.SubgroupanalysesshowedthatthedifferencewaslargeramongmenthanwomenandforpeoplewithMDscomparedtoMSDs.InGer‐many,theRBofhighmethodologicalquality,byStreibelt(50),foundthat88.4%ofthePTSLgrouphadreturnedtoworkat15monthsfollow‐upcomparedtoonly72.6%ofthecontrols(RelativeRisk[RR]=1.22,95%CI1.13–1.31).TherelativeriskofreturningtoworkwasgreaterinthePTSLgroupcomparedtoFTSL.ThegreatesteffectofPTSLonRTWwasobservedamongemployeeswhodidnotbelievethattheywouldgobacktoworkafterrehabilitation(74%inPTSLvs49%inFTSL).UnemploymentThreeRBsreportedonunemployment.InNorway,Markussenandcolleagues(49)ob‐servedthatPTSLpredictedariseintheemploymentprobabilitytwoyearsafterthestartofthesickleavebyaround16%(SemiparametricLeastSquares(SLS)estimate0.16,standarderror0.04),aftercontrollingforpatient/jobcovariatesandphysiciancharacteristics.Thisprobabilityincreasedupto20%whenspellsexceeding12weekswereanalyzed.InGermany,Streibelt(50)foundthatsick‐listedemployeesassignedtoPTSLhada60%lowerriskofunemploymentcomparedtopeerswhowereassignedto

  • 39 Results

    FTSL(RR0.41,95%CI0.26to0.65).Finally,TheFinnishRBpublishedbyViikari‐Jun‐turaandcolleagues(51)observedthatsick‐listedemployeeswhoreceivedPTSLspentlesstimeunemployedduringthe2‐yearfollow‐upcomparedtothosewhoreceivedFTSL.Thedifferenceinproportionswasabout1.8‐fold(3.2%inFTSLvs1.8%inPTSL).Thisdifferencewaslargeramongmenthanwomenandinworkersinmanufacturingcomparedtootherindustries.Insummary,oneRCTfromFinlandandallofthenineRBsthatmeasuredRTW,excepttheNorwegianRBbyLie(43),indicatedpositiveeffectsonRTWforemployeesonPTSLcomparedtoFTSL(15;49‐51).FavorableeffectsofPTSLcomparedtoFTSLonunem‐ploymentwerereportedbyallofthethreeRBsthatmeasuredthisoutcome(49‐51).Degreeofdisabilityandrehabilitationbenefits

    Therewerefivetypesofoutcomeswithregardtodegreeofdisabilityandrehabilitationbenefits:recurrenceofsickleaveforanycause,disability,productivityloss,disabilitypension,andallowanceofsocialwelfarebenefits.Wereportthestudyresultsforeachofthesefiveoutcomesseparatelybelow.RecurrenceofsickleaveforanycauseTheFinnishRCT(moderatemethodologicalquality)(25)indicatedthattimetofirstre‐currentsickleavewassimilarinthePTSLandFTSLgroups.However,thenumberofrecurrentsickleavesperpersonyearaftertheinitialsicknessabsenceperiodwasabout20%lowerinthePTSLgroupatone‐yearfollow‐up.DisabilityTheFinnishRCTandthreeRBsreportedondisability.Allhadmoderatemethodologicalquality.NodifferencesbetweenPTSLandFTSLregardingdisabilitywerereportedbytheFinnishRCTatone‐yearfollow‐up(47).ThethreeRBsreportingondisabilitywerefromNorway(twostudies)andFinland.Us‐ingNorwegiandata,Lie(43)foundnostatisticallysignificantdifferencesbetweenPTSLandFTSLonemployees’disabilityaftercontrollingforcovariates(i.e.,age,sex,diagno‐sis,timeinjob,andphysiciandata).Grødemandcolleagues(15)observedthelargestrecoveryamongemployeeswhohad

  • 40 Results

    4.5‐foldcomparedtotheFTSLgroup(7.9%versus1.8%);theoverallabsoluteriskdif‐ferencewas‐6.1%(95%CI‐7.1to‐4.9)(negativevalueindicatingincreaseinrisk).Higherriskswereseenamongwomen,theoldestemployees,andpeoplewithMSDscomparedwithMDs,andamongpeopleinthepublicsectorandhealthcareandsocialwork.Theseresultsremainedevenafteradjustingforresidualimbalanceinbaselinecovariates(e.g.,age,majorregion,employmentsector,socioeconomicstatus,andan‐nualgrossincome).Insum,oneRCTfromFinland(47)andoneRBfromNorway(43)foundnodifferencesbetweenPTSLandFTSLondisability,whereastwoNordicRBssuggestedpositiveef‐fectsofPTSLonemployees’disability(15;51),exceptwithregardtopartialdisabilityretirementintheFinnishRB(51).ProductivitylossOnlytheFinnishRCTreportedonproductivityloss.DatafromthisRCTofmoderatemethodologicalqualityfoundtherewerenosignificantdifferenceonproductivitylossbetweenPTSLandFTSLuptoone‐yearfollow‐up(regressioncoefficient‐0.6;95%CI‐9.1to7.9;p‐value=0.88).However,thiseffectbecamefavorableforPTSLafteradjust‐ingforbodymassindex,follow‐uptime,timesincebeginningofsymptoms(numberofelapseddays)andthebaselinemeasurebutdidnotreachstatisticalsignificance(re‐gressioncoefficient2.3;95%CI‐4.8to9.5;p‐value=0.52)(47).DisabilitypensionFourRBs,fromNorway,Finland,andGermany(twostudies),reportedondisabilitypension.ThreeoftheRBshadmoderatemethodologicalqualityandonehadhighmethodologicalquality.InNorway,Grødemandcolleagues(15)reportedthatthehigh‐estrateofreceivingdisabilitypensionwasobservedinemployeeswhowereon50%PTSL(12%attheterminationofsicknessbenefits,and19%twoyearslater),whereastheratevariedfrom7%upto13%attheendoftheobservationperiod(January2014)amongemployeeswhowereonFTSL.Thelowestlikelihoodofreceivingadisabilitypensionwasseeninthegroupof

  • 41 Results

    MDsthaninMSDs.NoassociationswereobservedbetweenPTSLandthetransitiontoanydisabilitypension(partialandfulldisabilitypensioncombined).Lastly,twoGermanRBsexamineddisabilitypension.Bethgeandcolleagues(17)re‐portedthatassigningPTSLtosick‐listedemployeesreducedtheirriskofreceivingadisabilitypensionby40%(HR=0.62,95%CI0.49–0.80)atone‐yearfollow‐up.Streibeltandcolleagues(50)reportedthatPTSLreducedtheriskofreceivingadisabilitypen‐sionby60%(RR=0.40,95%CI0.23‐0.70)comparedtoFTSLat15monthsfollow‐up.Thisstudyhadhighmethodologicalquality.Insum,acrossthefourRBs,theresultsfordisabilitypensionweremixed.TheNorwe‐gianRBsuggestedahigherrateofreceivingdisabilitypensionamongemployeesonPTSLcomparedtoFTSL(15),whereasthetwoGermanRBsreportedadecreasedriskofreceivingadisabilitypensioninemployeesonPTSL(17;50).TheFinnishstudy(24)foundthatPTSLreducedtheriskoffulldisabilitypensioncomparedtoFTSL,whereastheoppositeresultwasfoundforpartialdisabilitypension.AllowanceofsocialwelfarebenefitsFourRBs,fromNorway(threestudies)andGermany,reportedonallowanceofsocialwelfarebenefits.AllfouroftheseRBshadmoderatemethodologicalquality.InaNorwegiansetting,Markussenandcolleagues(49)foundthatPTSLwasassociatedwithfewersocialsecurityclaims(regressioncoefficientadjustedforpatient/jobco‐variatesandphysiciancharacteristics:‐79.7(standarderror11.8))inthe2‐yearperiodfollowingjustaftertheendoftheabsencespell.Thisreductionincreasedupto99daysreceivingbenefitswhenspellsexceeding12weekswereanalyzed.Lie(43)foundthattheprobabilityofreceivingsocialbenefitswasloweramongthoseonPTSLthanFTSL,althoughthedifferencesweresmallandnon‐significant(HRrangedfrom0.95at4weeksto0.80at28weeks)aftercontrollingforcovariates(i.e.,age,sex,diagnosis,timeinjob,anddoctor’sdata).Thesequentialanalysisdidnotrevealdifferencesbetweensecondandthirdsickleaves.Grødemandcolleagues(15)reportedthatallgroups,butnot50%(69%attermination,and40%twoyearsafter),and50%PTSL(59%attermination,and30%twoyearsafter).Theallowanceofsocialbenefitsremainedstableat10%inthe50%PTSL.TheGermanstudy,byBethgeandcolleagues(17),reportedthatemployeeswithPTSLreceivedfewerwelfarebenefitsduetosicknessabsenceandunemploymentuptotheendofthestudyperiodthanthoseinFTSL.Theaccumulatedtimeofreceivingsicknessbenefitswasreducedby52days(95%CI40–64days),short‐termunemploymentben‐efitsby58days(95%CI49–67days),andlong‐termunemploymentbenefitsby15days(95%CI10–20days)atone‐yearfollow‐up.

  • 42 Results

    Thus,allfourRBs(15;17;43;49)observedalowerallowanceofsocialwelfarebenefitsinPTSLcomparedtoFTSL.Health‐relatedoutcomes

    Onlytwooftheincludedstudiesreportedonhealth‐relatedoutcomes.ThiswastheFinnishRCT(25;47)andaGermanRB(50).TheRCThadmoderatemethodologicalqualityandtheRBhighmethodologicalquality.TheFinnishRCTincludedpeoplewithMSDswhiletheGermanstudyincludedpeoplewithMDs,primarilyaffectivedisorders.Below,wereporttheresultsforthesixhealth‐relatedoutcomesseparately.Thesewere:pain,self‐ratedgeneralhealthandhealth‐relatedqualityoflife,physicalandemotionalfunctioning,depression,anxiety,andworkingability.Pain(intensityandinterferencewithwork)ResultsfromtheFinnishRCT(47)showedreductionsinbothpainintensityandinter‐ferencewithworkinallgroupsduringthefirst8weeksandstabilizedthereafter.NodifferencesbetweenthePTSLandFTSLgroupswereobservedduringa12weeksfol‐low‐upperiodafteradjustingforbodymassindex,follow‐uptime,timesincebeginningofsymptoms(numberofelapseddays)andthebaselinemeasure.Thus,painintensity(≤3months)showedaregressioncoefficientof‐0.4(95%CI‐1.3to0.4;p=0.31);paininterferencewithwork(≤3months)‐0.7(95%CI‐1.6to0.3;p=0.15);paininterferencewithsleep(≤3months)‐0.12(95%CI‐0.9to0.7;p=0.77),andpainat1year‐0.2(95%CI‐0.7to0.4;p=0.48).Self‐ratedgeneralhealthandhealth‐relatedqualityoflifeTheFinnishtrial(47)foundthatemployeeswhoreceivedPTSLself‐reportedbettergeneralhealththanthoseintheFTSLgroup(regressioncoefficient0.5,95%CI‐0.0to1.0;p=0.07),andhigherhealth‐relatedqualityoflifeatone‐yearfollow‐up(regressioncoefficient‐0.5,95%CI‐0.9to‐0.01;p=0.02).Theseanalyseswereadjustedforbodymassindex,follow‐uptime,timesincebeginningofsymptoms(numberofelapseddays)andthebaselinemeasureoftheoutcome.PhysicalandemotionalfunctioningStreibeltandcolleagues’(50)resultsinGermanyshowedthatemployeeswhoreceivedPTSLhadahigherphysical(regressioncoefficient+7.9,p=0.01)andemotional(regres‐sioncoefficient+6.8,p=0.025)rolefunctioncomparedtothoseintheFTSLgroupat15monthsfollow‐up.DepressionNodifferenceswereobservedintheFinnishRCTbetweenthePTSLandFTSLgroupsonsick‐listedemployees’depressionsymptoms(47).Conversely,intheGermanRB(50),peopleinthePTSLgroupdidbetterthanpeopleintheFTSLgroup(regressionco‐efficient−0.6,95%CI−1.1to−0.1;p‐value=0.03)at15monthsfollow‐up.AnxietyDatafromtheGermanRB(50)showedthatpeopleinthePTSLgroupimprovedwithrespecttoanxietysymptoms(measuredwiththesametoolasdepression),compared

  • 43 Results

    toFTSL(regressioncoefficient−0.6,95%CI−1.1to−0.1;p‐value=0.03)at15monthsfollow‐up.WorkingabilityTheGermanRB(50)foundthatpeopleinthePTSLgroupdidbetterthanpeopleintheFTSLgroupwithrespecttoworkingability(regressioncoefficient0.1,95%CI‐0.1to0.3;p=0.05)at15monthsfollow‐up,butthedifferencewasnotstatisticallysignificant.Summaryofresultsacrosstheincludedstudies

    Table6summarizesthedirectionofresultsacrossthe13includedstudies(15out‐comes).Withregardtocausaleffects,theRCTpresentsthestrongeststudydesignandisthereforehighlighted.However,giventhestudy’ssmallsamplesize(n=62)firmcon‐clusionsabouttheeffectsofPTSLcannotbedrawn.

  • 44 Results

    Table6.Summaryofdirectionofresultsacrosstheincludedstudies(n=13)

    Legend:+favorstheintervention(partialsickleave,PTSL);‐favorsthecontrolgroup(full‐timesickleave,FTSL);=nodifferencebetweenPTSLandFTSL;*Randomized‐controlledtrial(RCT);#studyofhighmethodologicalquality,remainingstudieshadmoderatemethodologicalquality.Emptycellmeansthestudydidnotexaminetheoutcome.

    Outcomemeasure

    André

    n2012(4

    8)

    André

    n2014(

    28)

    Bethg

    e2016 (1

    7)

    Grødem

    2015 (15

    )

    Høgel

    und2

    010 (

    52)

    Kaust

    o2012 (2

    4)

    Kaust

    o2014 (1

    )

    Lie20

    14 (43

    )

    Marku

    ssen2

    012 (4

    9)

    Nosse

    n2013 (

    44)

    Shiri2013(47)*

    Viikari‐Juntura2012(25)*

    Streib

    elt20

    17 (50

    )#

    Viikari‐Jun

    tura2

    017 (

    51)

    Sicknessabsence + + + + WorkparticipationReturn‐to‐work + + + + + ‐ + + + +Unemployment + + +DegreeofdisabilityandrehabilitationbenefitsRecurrenceofsickleaveforanycause = Disability + = = +Productivityloss = Disabilitypension + ‐ + + Allowanceofsocialwelfarebenefits + + + + Health‐relatedoutcomesPain(intensityandinterferencewithwork)

    = Self‐ratedgeneralhealth + Health‐relatedqualityoflife + Physicalandemotionalfunctioning

    + Depression = + Anxiety + Workingability +

  • 45 Discussion

    Discussion

    Mainfindings

    Todate,12RBs(fromtheNordiccountriesandGermany)andoneRCTfromFinlandhaveinvestigatedtheeffectsofPTSLcomparedtoFTSLinmorethan2.74millionsick‐listedemployees.Allstudieshadmoderatemethodologicalquality,exceptoneRBs,whichhadhighmethodologicalquality.Thisindicatesconsistenthighinternalvalidityintheresearchmethodsused,butitisimportanttostressthatRBshavelimitedcapac‐itytodetectcausaleffects.TheFinnishRCTstudiedemployeeswhoweresick‐listedduetoMSDs,whiletheRBsincludedemployeeswithprimarilyeitherMSDsorMDs.Whilefirmconclusionscannotbedrawn,theRCTandtheRBssuggestedPTSLmaybeassociatedwithshortersicknessabsenceandhigherworkparticipation.TheFinnishRCTreportedthatemployeeswithPTSLexperiencedbettergeneralhealthandqualityoflifecomparedtothoseonFTSL.However,itdidnotfindstatisticaldifferencesbe‐tweenPTSLandFTSLonsickleaverecurrence,employees’productivityloss,andpain.TheRBsindicatedalowerprobabilityforpeopleonPTSLofreceivingbothdisabilitypensionandallowancebenefits,disability,aswellasbetterscoresonphysical‐andemotionalfunctioning,anxiety,depression,andworkingability.TheresultsregardingsicknessabsenceandworkparticipationaresupportedbyaNor‐wegianRB(26),whichusedthesameNAVdatasetasLie(43)andpartiallyMarkussenandcolleagues(49),bothincludedinthisreview.Kannandcolleagues(26)demon‐stratedthatanincreaseintherateofuseofPTSLinamunicipalityofonepercentagepoint(e.g.from13%to14%)wasassociatedwithareductioninthesicknessabsencerateof1.79%,andshortersicknessabsenceduration.Thatis,whentheproportionofdaysofsickleavethatweregradedincreasedbyonepercentagepoint,therateofsick‐nessabsencedecreasedbyabouttwopercent.Thus,theresearchersconcludedthatgreateruseofPTSLcaninsubsequentmonthsleadtoareductioninthesicknessab‐sencerate,durationofsicknessabsence,andnumberofindividualsonsickleave.Asimilaranalysiswithdatafrom2000to2011reachedthesameconclusions,althoughtheassociationsinthisanalysiswereweaker(46).Oursystematicmappingreviewidentifiedsomegapswithregardtothedifferentout‐comesmeasuredacrossstudies.Workparticipationwasthemostcommonoutcomere‐ported,measuredintenstudies,followedbysicknessabsenceduration,disability,disa‐bilitypensionandallowanceofsocialwelfarebenefits,withfourstudieseach.Incon‐trast,therecurrenceofsickleavewasonlymeasuredinonestudy(theRCT).Therewas

  • 46 Discussion

    animportantlackofknowledgeregardingtheeffectofPTSLforhealth‐relatedout‐comes,asonlytwostudiesaddressedthisissue(table6).Further,itshouldbenotedthattheoverwhelmingmajorityofthestudypopulationswereadultssufferingfrommusculoskeletal‐ormentaldisorders.Noneofthestudiesspecificallyincludedpeopleonsickleaveforotherreasons.ItispossiblethateffectsofPTSLarerelatedtodiagno‐sis.Forexample,Høgelundandcolleagues(29),whocombinedsurveyandregisterdataonabout850Danishworkers,foundthatPTSLhadnoeffectonthedurationuntilre‐turningtoregularworkinghoursforemployeeswithMDs,butsignificantlyreducedthedurationuntilreturningtoregularworkinghoursforemployeeswithotherdisorders.Generalizabilityandstrengthoffindings

    TheOECDhasreportedthattheinsufficientlaborforceparticipationamongpeoplewithhealthissuesanddisability,theirlowincome,andthehighcostsofsicknessanddisabilitybenefitschemesrepresentaseriousproblemforgovernments,andtheuseofPTSLmightbeasuitablemeasuretocounteractthem(2).Animportantpointinfavorofthegeneralizabilityofthefavorablefindingsfromthestudiesincludedinthissys‐tematicmappingreviewisthepredominanceofNordicstudies.Elevenofthe13studieswerefromaNordiccountry,includingfourfromNorway.Further,aconsiderablegroupofemployeesintheincludedstudiesweresick‐listedduetoMSDsandMDs.TheformerrepresentsthemainoccupationaldiseasessufferedbyEuropeanworkers,accordingtotheEuropeanObservatoryofWorkingLife(55).InNorway,MSDsareamongthemostcommonreasonsforconsultationtoGPsandemergencyprimaryhealthcare,andrep‐resentthemajorityofdaytreatmentsinthenationalhealthsystem(56).Moreover,withregardtoMDs,datafromtheSwedishSocialSecurityAgencyshowthatMDsac‐countedforaroundhalfofallsickleavecasesamongwomenand40%ofallcasesamongmeninDecember2016.In2016,PTSLrepresentedaround30%ofallsickleavecasesandwasespeciallyhighamongpeoplewithMDs(14).AllRBsuseddatafromstructurednationwideregistries.TheNorwegianstudiesuseddatafromNAV,whichmightfacilitatetheformulationofpublicpoliciesandfurtherre‐searchoftheuseofPTSLforcontrollingsicknessabsence.WenotethatthetwoGer‐manstudiesconcernedatotalofabout4500employees–andthereforecontributeslesstotheoverallfindings–whocompletedarehabilitationprogram,whichismanda‐toryinGermany,butnotstandardpracticeinNorwaywherepeoplegenerallystartPTSLearlyinthecourseofthesicknessabsence.InGermany,PTSLisgrantedwhenthesick‐listedemployeehascompletedtherehabilitationprogrambutisstillunabletoperformfullduties,anditisapprovedbytherehabilitationphysicianandthesocialworkerwithconsentfromtheemployer,thepatient,thegeneralpractitionerandtheoccupationalphysician(57‐59).Aspartoftheschemedevelopedbytherehabilitationphysician,thesick‐listedemployeestartsworkingforatleasttwohours/dayandgrad‐uallyincreasestheworkingtime(58;59).

  • 47 Discussion

    Strengthofevidence

    Thetwotypesofstudydesignsandresearchapproachesthatrepresentthebodyofevi‐denceontheeffectsofPTSLversusFTSLmeritsomediscussionintermsofthepossi‐bilitytodrawcausalinferencesfromtheirfindings.TheFinnishRCTrepresentsthestrongeststudydesigntoanswertheresearchquestionthatguidedthissystematicmappingreview,asitiswidelyacceptedthatwell‐conductedprospective,experimentalstudieshavethegreatestcapacitytodetectcausaleffectsduetotheirpossibilitytoran‐domlyassignindividualstodifferentexposures,andthereforeensurethatgroupsaresimilar(60).Essentialsupportforthisstatementmaybebasedonthecounterfactualapproachforevaluatingcausalinferenceinepidemiology,whicharguesthatahighcomparabilitybetweenexposedandunexposedindividualsisneededtoestimateanycausaleffect,asitisnotpossibletoobserveanindividual’soutcomebothatthetimewhenheisexposedandatthesametimewhenheisnot(61‐63).Nevertheless,randomassignmentisnotsufficienttoensurehighvalidityofexperimentalstudies,asalsothesemayhavelimitationsthatcanleadtobiasedestimatesofcausaleffects(e.g.,lowadherencetotheintervention,highattrition,andoutcomemeasurementerrors)(40‐42;64;65).Additionally,ethicalconstraintsforconductingexperimentalresearchareacommonconcernthatturnsthefocustoobservationaldata.Inspiteofthepreferenceforexperimentalstudieswhenaddressingquestionsabouteffect,andasnoticedinourfindings,thebodyofevidenceaboutthebenefits/draw‐backsofPTSLcomparedtoFTSLismostlyrepresentedbyRBsthatuseobservationaldataderivedfromnationwideregistries.Whileabroaddiscussionaboutcausalinfer‐encegoesbeyondthescopeofthissystematicmappingreview,weofferafewnotesaboutthecapacitytodrawcausalinferencesfromobservationaldata.Observationalstudiesestimatedifferencesonacertainoutcome“X”amongindividualswithdifferentlevelsofanexposure“Y”(63;66),andnumerousmethodologicalapproachesarede‐scribedintheliteraturetoestimatecausaleffectsbasedontheirdata(62;63;66).Someofthemajorconstraintstodetectcausalinferencesinobservationaldataaretiedtothesusceptibilityofselectionbiasduetoanon‐randomdistributionoftheexposure,thepossibilityofconfounding,reversecausation,andtheoptiontoseeoutcomedatabe‐foretheallocationoftheindividualstothegroupsofstudy(62;67).Outcomedataanal‐ysisisnotprecededbytheallocationoftheindividualstodifferentexposurelevels,thusanequaldistributionofcovariatesbetweenexposedandunexposedindividualsmaynotbeensured(66;6