Characteristics of patients with knee and ankle symptoms accessing physiotherapy: self-referral vs general practitioner's referral

Physiotherapyxxx(2020)xxx–xxx

Characteristics

of

patients

with

knee

and

ankle

symptoms

accessing

physiotherapy:

self-referral

vs

general

practitioner’s

referral

N.E.

Lankhorst

,

J.A.

Barten

,

R.

Meerhof

,

S.M.A.

Bierma-Zeinstra

,

M.

van

Middelkoop

a_{DepartmentofGeneralPractice,ErasmusMC,UniversityMedicalCentre,Rotterdam,TheNetherlands} b_{NetherlandsInstituteforHealthServicesResearch,Utrecht,TheNetherlands}

Abstract

Objectives Todeterminepatientcharacteristics,frequencyofuse,typeofsymptomsandtreatmentoutcomesinpatientswithkneeorankle symptoms,separately,forpatientsreferredbytheirgeneralpractitioner(GP)andpatientswhoself-referred.

Design Longitudinalstudy.

Setting Dutchprimarycarephysicaltherapypractices.

Participants Allpatientswhovisitedaphysiotherapistwithkneeoranklesymptomsbetween2006and2012.

Method DatawerecollectedfromtheNIVELPrimaryCareDatabase.Themodeofaccess(self-referredorGP-referred)wasdeterminedfor allpatients.Foranalyses,descriptivestatistics,unpairedt-tests,Chi-squaretestandlogisticregressionanalyseswereapplied.

Results Thestudyincluded6179patientswithkneeoranklesymptoms.Theuseofself-referralincreasedfrom26%in2006to56%in 2012,andstabilisedbetween2010and2012.Self-referredpatientswereyounger,hadahighereducationlevelandhadashorterdurationof symptomscomparedwithGP-referredpatients.Self-referredpatientshadfewertreatmentsessionsthanGP-referredpatients.

Conclusions Amongpatientswithkneeoranklesymptoms,youngerpatients,andthosewithahighereducationlevel,ashorterdurationof symptomsandrecurrentsymptomsweremorelikelytoself-referthanotherpatients.Self-referredpatientshadfewertreatmentsessions.After 2009,thefrequencyofself-referralstophysicaltherapistsstabilised.Futurestudiesshouldexaminetheeffectivenessofphysicaltherapyfor patientswhoself-referwithacutekneeandanklesymptoms.

©2017CharteredSocietyofPhysiotherapy.PublishedbyElsevierLtd.Allrightsreserved.

Keywords:Directaccess;Physiotherapy;Knee;Ankle

Introduction

Musculoskeletalsymptoms(e.g.back,shoulder,kneeand ankle)arecommoninprimarycare,andaccountfor approx-imately12%of allconsultationsingeneralpracticeinThe Netherlands[1,2].Kneeandanklesymptomsarecommon, withincidenceratesof10.9to13.7per1000personsand3.3 per1000persons,respectively[2,3].

∗_{Correspondingauthor at:Department ofGeneral Practice, Erasmus}

MC,UniversityMedicalCentre,POBox2040,3000CARotterdam,The Netherlands.Fax:+31107032127.

E-mailaddress:m.vanmiddelkoop@erasmusmc.nl (M.vanMiddelkoop).

Variousnon-surgical treatmentoptions areavailablefor kneeandanklesymptoms,andmainlycompriseadvicetorest anduseanalgesicmedicationforacutetraumaticsymptoms

[4,5].Physiotherapyisalsoconsideredfornon-traumaticand chronicknee/anklesymptoms,includingchronicsymptoms duetotrauma[4,6,7].InTheNetherlands,although physio-therapyisnot stronglyrecommendedinthe guidelinesfor generalpractitioners(GPs),thereisevidencethatsupervised exercisetherapycanresultinpainreductionandfunctional improvementcomparedwithusualcareinpatientswith non-traumatic kneesymptoms,traumatickneeinjuryandankle injury[4,8–10].

https://doi.org/10.1016/j.physio.2017.03.008

OneresultofthenewhealthcaresystemintroducedinThe Netherlandsin2006isthatpatientsnolongerneedaformal referralbyaGP(orotherphysician)toconsulta physiothera-pist.Thischangeledtoanincreaseinthenumberofpatients withmusculoskeletalcomplaintsconsultingphysiotherapists throughself-referral:inthefirstyearofintroduction(2006), 32%ofpatientsconsultedphysiotherapistswithouta refer-ralfromaGP,andthisincreasedto51%ofpatientsin2015

[11,12].

Three previous studies in The Netherlands comparing thecharacteristicsofpatientswhovisitedaphysiotherapist followingaGPreferralorwhoself-referredfoundthat self-referredpatientshadahighereducationlevelandashorter durationofsymptoms[13–15].Thefirststudywasconducted immediately after the implementation of self-referral and hadashorterdurationoffollow-up[13],whereastheother two studiesevaluatedpatient characteristicsbetween 2006 and2009[14]andbetween2006and2010[15].However, the studiesincluded patients withback pain alone [14]or patientswithdiversetypesofmusculoskeletaldisease(e.g. back,neck,shoulder)[15],andpatientswithkneeandankle symptomsmaydifferfromthesepatients.Patientswithknee andanklesymptomsaremorelikelytobedistributedbetween differentagegroups(includingyoungandolderpatients),and willprobably includeapercentageoftraumatic symptoms thatcouldimpactthechoiceofcare[4,5].

Todate,nostudieshaveexaminedtheassociationbetween patient characteristics and the mode of access to physio-therapyamong patients withknee or ankle symptoms. As knee and ankle complaints are among the most common typesofsymptomsinphysiotherapypractice,physiotherapy isconsideredastreatmentforbothinprimarycare,andboth representtraumaticandnon-traumaticpatients,theprimary aims ofthisstudy were: (i)toestablish the distributionin mode of access of patients with knee or ankle symptoms between2006and2012;and(ii)toinvestigatethedifferences inpatientandtreatmentcharacteristicsbetweenGP-referred patientsandself-referredpatients,andtoidentifythe charac-teristicsassociatedwithself-referral.

Methods

Data were collected from the NIVEL Primary Care Database (NPCD),formerly known as the National Infor-mation Service for Allied Health Care. NPCD consists of longitudinal data, collected by extraction of routinely recordeddatainthe healthcareprovider’selectronichealth recordsystem.ThedatacollectedbyNPCDarefora repre-sentativesampleof theDutchpopulation,andincludedata on health problems and treatment. GPs, physiotherapists, exercise therapists, dieticians, primary care psychologists and GP out-of-hours services contribute to NPCD. In the currentstudy,onlydataandregistrationsofpatientswho vis-itedaphysiotherapistwereused.Thephysiotherapistswere selectedfromthosewhoworkextramurally(inthe

commu-nity)inTheNetherlands,andarenationallyrepresentativeby regionaldistribution.MoredetailedinformationaboutNPCD hasbeenpublishedpreviously[13–15].

Studypopulation

Thestudypopulationconsistedofallpatientswhovisited aphysiotherapistwithkneeoranklesymptomsbetween2006 and2012.Kneeandanklesymptoms wereidentified using theInternationalClassificationofPrimaryCare(ICPC)[16]. EachtreatmentepisoderepresentedinNPCDwaslinkedto oneormoreICPCcodes.InthecaseofGPreferral,the writ-ten record made by the referring GP was recoded; in the caseof self-referral,the mainhealthproblemwas recoded into an ICPC code. A digital self-learning module devel-oped over the last 10 yearsof registration recognises text wordsandcodes90%ofcasesautomaticallyintoICPCcodes. Theadditional10%areenteredmanuallyintothedatabase. Aresearchassistantmonitoredandverifiedthisdigitalised recoding process.ThefollowingICPCcodes wereusedto select patients with knee symptoms: knee symptoms and complaints(L15),kneesprain/distortion(L78),osteoarthritis of the knee (L90), Osgood–Schlatter/osteochondritis dis-secans (L94), acute meniscal or ligamental injury (L96), chronic internal knee trauma (L97), patellofemoral syn-drome(L99.07),corpusliberum(L99.10)andpseudarthrosis (L99.11).Toselectpatientswithanklesymptoms,the follow-ingcodeswereused:anklesymptomsandsymptoms(L16) and ankle sprain/distortion (L77).The typesof symptoms were divided into traumatic andnon-traumatic symptoms. Knee sprain/distortion (L78) and meniscal or ligamental injury(L96)wereusedtodescribetraumatickneesymptoms. Anklesprain/distortion(L77)wasusedtodefinetraumatic anklesymptoms.

InNCPD,eachepisodeofkneeoranklesymptomswas registeredseparately.Anepisodewasdefinedasthe occur-rence ofkneeoranklesymptomsfromthestartofthefirst consultationtotheendoftreatmentbetween2006and2012; to avoid double counting of individual patients, only the first episode of carewas included for patients with multi-ple episodes.Patientswereexcludedifthemode ofaccess wasnotreported,iftheyhadbothkneeandanklesymptoms inthesameepisode,oriftheywerereferredbyaphysician otherthanaGP.

Ethical approval was not requiredbecause NPCD does notfallwithinthescopeoftheMedicalResearchInvolving HumanSubjectsAct.Datawerecollectedanonymouslyand patients were informed about the research by posters and leafletsinpracticewaitingrooms.Thestudyadheredtothe tenetsoftheDeclarationofHelsinki.

Measurements

Between2006and2012,thefollowingdatawerecollected foreachepisode.

Patientcharacteristics

Age(years),sex,urbanisation(urban/rural),patient iden-tification number, and educational level in patients aged >16years.Educationlevelwascategorisedaslower educa-tion (primary school,lower vocational educationor lower secondaryeducation),middleeducation(intermediate voca-tionaleducationorintermediate/highersecondaryeducation) andhighereducation(highervocationaleducationand uni-versity).

Referral

Diagnosis(basedonICPCcode)andmodeofaccess (self-referralorGPreferral).

Healthproblem

Recurrenceofkneeoranklesymptoms(visitinga physio-therapistwiththesamesymptomsmorethan3monthsafter the termination of care for the first episode) (yes/no) and durationofsymptoms(≤7days,1weekto1month,1month to12months,≥12months).

Treatment

Numberofvisitstothephysiotherapistforcurrentepisode, anddurationoftreatmentepisode(totalnumberofdaysthat thepatientvisitedaphysiotherapist).

Evaluation

Self-reported reason for termination of care (goals achieved,noinsurance,terminatedbypatient,terminatedby physiotherapistorterminatedbyreferrer)andself-reported resultsonthebasisofthetreatment goalsetatthestartof treatment(achieved,partiallyachievedornotachieved).

Dataanalyses

Dataonkneeandankle symptomswereanalysed sepa-rately.AnalyseswereperformedusingSTATAVersion13.0 (StataCorp,CollegeStation,TX,USA).

Patient characteristics, treatment outcomes and evalua-tionoftreatmentwereanalysedusing descriptivestatistics. DifferencesbetweenGPreferralandself-referralwere ana-lysedusingunpairedt-testsinthecaseofnormallydistributed data;Chi-squaredtestswereusedinthecaseofnon-normally distributedanddiscretedata.

Amultivariatelogistic regressionmodel withbackward stepwise selection [P (in)0.05,P (out) 0.10] was used to studytheassociationbetweenmodeofaccessto physiother-apyandpatientcharacteristicsinpatientswithkneeorankle symptoms.Thefollowingcharacteristicswereincludedinthe analysis:sex, age,diagnosis,educationlevel, urbanisation, durationofsymptomsandrecurrentsymptoms.Associations were presented as odd ratios (ORs)with 95% confidence intervals(95%CI).Tocheckwhethermissingdatabiasedthe results,anadditionalmultivariatelogisticregressionanalysis wasperformed;thisincludeddummyvariables(withvariable

Fig.1.Patientflowchart.PT,physiotherapist.

‘missing’)andcomparedtheoutcomesofbothmodels(with andwithoutthedummyvariables).

Results

Studypopulation

Fromtheinitial76806patientswhovisiteda physiothera-pistbetween2006and2012,6179patientswithkneeorankle symptomswereselectedforinclusioninthisstudy(Fig.1). Ofthese,themeanagewas41(range4 to101)years,and 2869(46%) weremen. In total,4688(76%) patients were diagnosedwithakneecomplaintand1491(24%)patients werediagnosedwithananklecomplaint(Table1).Patients withmissingdatawerehadanoldermeanageandmoreoften hadinjuriesoftraumaticorigincomparedwithpatientswith completedata.

The majority of patients with knee or ankle symptoms (62%)visitedaphysiotherapistfollowingreferralbyaGP.

Table1

Characteristicsofpatientsvisitingaphysiotherapistduetokneeoranklesymptoms,bymodeofaccess.

Kneesymptoms Anklesymptoms

GPreferral Self-referral P-value GPreferral Self-referral P-value n=2949 n=1739 n=865 n=626

n(%) n(%) n(%) n(%)

Sex:male 1343(46) 840(48) 0.07 389(45) 297(47) 0.34

Missing 2(0) 0(0) 0(0) 0(0)

Age(years),mean(SD) 44(21) 41(20) <0.001 34(20) 33(18) 0.15

Agegroup(years) <0.001 0.13

<15 214(7) 111(6) 144(17) 88(14) 15to24 510(17) 389(22) 215(25) 181(29) 25to34 308(10) 248(14) 117(14) 89(14) 35to44 435(15) 242(14) 110(13) 95(15) 45to54 517(18) 274(16) 124(14) 85(14) 55to64 439(15) 247(14) 81(9) 55(9) 65to74 283(10) 133(8) 53(6) 24(4) ≥75 243(8) 95(6) 21(2) 9(1) Educationlevel <0.001 <0.001 Lower 762(26) 313(18) 193(22) 81(13) Middle 732(25) 415(24) 180(21) 151(24) Higher 491(17) 547(32) 152(18) 178(28) Missing 964(33) 464(27) 340(39) 216(35) Urbanisation <0.001 0.32 Urban 1387(47) 846(49) 373(43) 273(44) Rural 1389(47) 678(39) 436(50) 286(46) Missing 173(6) 215(12) 56(7) 67(11)

Durationofsymptoms(months) <0.001 0.007

<1 936(32) 744(43) 460(53) 386(62)

1to3 824(28) 426(25) 201(23) 119(19)

3to12 637(22) 296(17) 114(13) 62(10)

>12 482(16) 218(13) 74(9) 44(7)

Missing 70(2) 55(3) 16(2) 15(2)

Recurrentsymptoms:yes 731(25) 450(26) 0.58 197(23) 177(28) 0.006

Missing 71(2) 58(3) 13(2) 17(3) Typeofsymptoms 0.43 <0.001 Traumatic 196(7) 126(7) 570(66) 354(57) Non-traumatic 2753(93) 1613(93) 295(34) 272(44) GP,generalpractitioner. Table2

Specificationofkneeandanklesymptoms,separately,foreachmodeofaccess.

Diagnosis(ICPCcode) Totalpopulation GPreferral Self-referral P-value

n(%) n(%) n(%) Overall

Kneesymptoms 4688 2949 1739 <0.001

Kneesymptoms/complaints(L15) 3509(75) 2071(70) 1438(83)

Sprainedknee(L78) 240(5) 145(5) 95(6)

Osteoarthritisoftheknee(L90) 382(8) 295(10) 87(5)

Osgood–Schlatter(L94.02) 28(1) 23(1) 5(0)

Acutemeniscalinjury(L96) 82(2) 51(2) 31(2)

Chronicinternaltraumaknee(L97) 14(0) 12(0) 2(0)

Patellofemoralsyndrome(L99.07) 433(9) 352(12) 81(5)

Anklesymptoms 1491 865 626 <0.001

Anklesymptoms/complaints(L16) 567(38) 295(34) 272(43)

Sprainedankle(L77) 924(62) 570(66) 354(57)

ICPC,InternationalClassificationofPrimaryCare;GP,generalpractitioner.

Ofall patients withkneesymptoms, 75% were diagnosed with ‘knee symptoms/complaints (L15)’. Other common knee diagnoses were‘patellofemoral syndrome’ (9%) and ‘osteoarthritisofthe knee’(8%).Themost commonankle symptom (62%) was ‘sprained ankle (L77)’. The major-ityofpatientsdiagnosedwith‘kneesymptoms/complaints’ were self-referrals. Of the patients with ankle complaints whoself-referred,themostcommondiagnosiswas‘sprained ankle’ (57%). However, of all patients with ankle com-plaints,the mostcommonmodeof accesswasGPreferral (Table2).

Timetrendsinmodeofaccess

Theproportionofself-referredpatientswithkneeorankle symptoms increased from 26% in 2006 to 56% in 2012 (Fig.2).Thisincreasewasmainlyobservedbetween2006 and2009.

Characteristicsofpatientswithkneesymptoms

Compared with GP-referred patients, self-referred patientswithkneesymptomswereyounger(41vs44years, respectively), had a higher education level, lived in more urbanised areas, and had a shorter duration of symptoms (Table1).

Furthermore, patients with knee symptoms who self-referredhadfewerphysiotherapysessionsandtreatmentdays comparedwithpatientswhowerereferredbyaGP:6.3vs 9.1sessionsand49.7vs60.2days,respectively(Table3).

Multivariateregressionanalysesshowedthatpatientswith kneesymptomsweremorelikelytoself-referiftheyhada highereducationlevel(OR2.65;95%CI2.19to3.20),and werelesslikelytoself-referiftheywereaged25to44years (OR0.72;95%CI0.58to0.89)or45to64years(OR0.67; 95%CI0.54to0.82)(comparedwiththereferencecategory aged<25 years),andhadalongerduration(>1month) of symptoms(OR0.57to0.65)(Table4).

Characteristicsofpatientswithanklesymptoms

Patients with ankle symptoms who self-referred had a higher level of education, lived in more urbanised areas, reported a shorter duration of symptoms, and were more likely to report recurrent symptoms compared with GP-referred patients. Patients referred by a GP had more traumatic symptoms compared with patients who self-referred(Table1).

Moreover,patientswithanklesymptomswhoself-referred hadfewertreatmentsessionsthanpatientswhowerereferred byaGP(5.6vs6.7sessions)(Table3).

Similartopatientswithkneesymptoms,multivariate anal-ysesshowedthat patientswithankle symptomsweremore likelytoself-referiftheyhadahighereducationlevel(OR 2.74;95%CI1.94to3.86),andwerelesslikelytoself-refer iftheywereaged25to44years(OR0.72;95%CI0.52to

0.99)or>65years(OR0.53;95%CI0.29to0.98)(compared withthereferencecategoryaged<25years),hadtraumatic anklesymptoms(sprainedankle)(OR0.54;95%CI0.40to 0.72)andhadalongerduration(>1months)ofsymptoms (OR0.54to0.70)(Table4).

Discussion

This study examined differences in characteristics betweenpatientswithkneeoranklesymptomswhovisited aphysiotherapistfollowingreferralbyaGPcomparedwith patientswhoself-referred.In2006,26%ofallpatientswho visitedaphysiotherapistchosetoself-refer,andthisincreased to56%ofallpatientsin2012;thisgrowthstabilisedbetween 2010and2012.Furthermore,youngerpatientswithahigher educationlevelandashorterdurationofsymptomsweremore likelytoself-refer.

Itisreportedthat olderpatientsandthosewithalonger durationofsymptomsareathigherriskofcomorbidityand morecomplexandchronicsymptoms,resultinginmorevisits toaGP[17–19].Thisissupportedbythecurrentfindingthat patientswerelesslikelytoself-referiftheysufferedfroma diagnosis withhigherprevalenceamongolderpatientsand ofamorechronicnature(e.g.osteoarthritis).

Similartootherstudies,thepatientsinthisstudywitha shorterduration(<1month)ofsymptomsweremorelikelyto referthemselvestoaphysiotherapist[13,14].Ithasrecently beenshownthat87%ofself-referredpatientswith muscu-loskeletalsymptoms whovisitedaphysiotherapistwerein fact treatedbyaphysiotherapist afterthe initialvisit[15]. It therefore seemsthat a large percentageof self-referred patientswhoseekhelpfortheirmusculoskeletalsymptoms (including knee and ankle symptoms) are actually treated bytheirphysiotherapist.GPsaregenerallyregardedas gate-keepers of careandmay beless likely toreferpatients to aphysiotherapist when there isdoubt regarding the effec-tivenessoftreatment.Fromthisperspective,itisremarkable that evidence on the (cost-)effectiveness of physiotherapy foracutekneeandanklesymptoms(durationofsymptoms <1 month) is lacking. Also, the present study found that the patient characteristics associated with self-referral are comparablewiththe prognosticfactorsfor betterrecovery rates inbothknee andanklesymptoms reportedbyothers (e.g.youngerage,shorterdurationofsymptoms,higherlevel ofeducation)[20,21].Thismayindicatethatthesepatients are more likelyto haveapositive outcome, withor with-outtreatmentfromaphysiotherapist.Therefore,information onprognosticfactorsforrecoveryandapositivetreatment outcome would bevaluable for physiotherapistsfor future decision-makingatfirstvisit(e.g.totreat,refer,orwaitand see).

Patientswithkneeoranklesymptoms whoself-referred toaphysiotherapisthadfewertreatmentsessionsanddaysof treatmentcomparedwithpatientswhowerereferredbyaGP. Theseresultsareconsistentwithotherstudiesthatcompared

Fig.2.Distributionofthemodeofaccessinpatientswithkneeoranklesymptomsbetween2006and2012.Darkerbarsrepresentself-referredpatients,and lighterbarsrepresentpatientsreferredbyageneralpractitioner.

Table3

Treatmentcharacteristicsofpatientsvisitingaphysiotherapistduetokneeoranklesymptoms,separately,bymodeofaccess.

Kneesymptoms Anklesymptoms

GPreferral n=2949 Self-referral n=1739 P-value GPreferral n=865 Self-referral n=626 P-value Treatmentsessionsa

Numberofsessions,mean(SD)* 9.1(11.2) 6.3(7.14) <0.001 6.7(7.4) 5.6(6.1) 0.006

Durationoftreatmentepisodeindays,mean(SD) 60.2(98.6) 49.7(74.4) <0.001 43.9(60.7) 41.7(58.4) 0.52

Goalsaccomplished?b _{0.17 0.25}

Goalsaccomplished 979(33.2) 541(31.1) 400(46.2) 218(34.8)

Goalspartiallyaccomplished 358(12.1) 196(11.3) 64(7.4) 49(7.8)

Goalsnotaccomplished 121(4.1) 48(2.8) 17(2.0) 11(1.8)

Reasonforterminationofcarec _{0.01 0.77}

Goalsaccomplished 1148(39.0) 634(36.5) 439(50.8) 260(41.5) Nocompensationinsurance 29(1.0) 7(0.4) 5(0.6) 2(0.3) Terminatedbypatient 94(3.2) 28(1.6) 12(1.4) 9(1.4) Terminatedbytherapist 106(3.6) 48(2.8) 14(1.6) 10(1.6) Terminatedbyreferrer 8(0.3) 2(0.1) 5(0.6) 1(0.2) Missing

GP,generalpractitioner;SD,standarddeviation. Missingdata:a_n=772;b_n=3177;c_n=3318.

differencesintreatmentcharacteristicsbetweenGP-referred andself-referredpatients[13,22–25].

Althoughpatientswhovisitedaphysiotherapistwithout referral by a GP had fewer treatment sessions compared withpatientswhowerereferredbyaGP,itremainsunclear whetherthedifferenceinthe numberoftreatment sessions isduetothemodeofaccessorotherpatientcharacteristics. Therefore,basedonthedatafromthisstudy,noconclusions canbedrawnaboutthecost-effectivenessofself-referral.

Strengthsandlimitations

Thedifferencesinpatientcharacteristicsbetweenpatients whoself-referredandpatientsreferredbyaGParein agree-mentwithtwoearlierstudiesconductedinTheNetherlands

[13,14]. Thesestudiesalso foundthat education level and durationofthecomplaintwereassociatedwithself-referral. However,thepresentstudyalsoanalysedtheeffectsofthe

availabilityof self-referralover alonger periodof timein specific musculoskeletal symptoms (i.e. knee and ankle). Therefore, the present study provides information on dif-ferencesbetweenpatientswithtraumaticandnon-traumatic kneeandanklesymptoms,aswellasmoreinsightintothe effectsoftheimplementationofself-referral6yearsafterits implementationinthisspecificpatientgroup.

All diagnosesinthe presentstudywere basedonICPC codes[16].Thesecodeswereallocatedbyacomputer pro-gram developed by NIVEL. Although all diagnoses were providedbythesamecomputerprogram,some methodolog-icaldifferencesmayexistintheallocationofthediagnosis betweenpatientswithandwithoutaGPreferral.Inpatients usingself-referral,alldetailsonthediagnosiscamefromthe physiotherapist’s record, whereasfor GP-referred patients, theexactwordingoftheGP’sreferralwasrecoded.Duetothis methodologicaldifference,theprevalenceratespresentedin thisstudyshouldbeinterpretedwithcaution.

Table4

Multivariatelogisticregressionforfactorsassociatedwiththeuseofself-referralinpatientswithkneeoranklesymptoms.

Kneesymptoms Anklesymptoms

Oddsratio(95%CI) P-value Oddsratio(95%CI) P-value

Agegroup(years)

<25 1.00(reference) 1.00(reference)

25to44 0.72(0.58to0.89) 0.002 0.72(0.52to0.99) 0.05

45to64 0.67(0.54to0.82) <0.001 0.72(0.51to1.02) 0.07

>65 0.80(0.61to1.05) 0.11 0.53(0.29to0.98) 0.04

Disorder

Kneesymptoms/complaints(L15) 1.00(reference) 0.004

Sprainedknee(L78) 0.61(0.43to0.86) 0.001

Osteoarthritisoftheknee(L90) 0.58(0.42to0.80) 0.51

Osgood-Schlatter(L94.02) 0.46(0.05to4.52) 0.18

Acutemeniscalinjury(L96) 0.71(0.43to1.17) 0.06

Chronicinternaltraumaknee(L97) 0.23(0.05to1.07) <0.001 Patellofemoralsyndrome(L99.07) 0.28(0.20to0.39)

Anklesymptoms/complaints(L16) 1.00(reference)

Sprainedankle(L77) 0.54(0.40to0.72) <0.001

Educationlevel

Lower 1.00(reference) 1.00(reference)

Middle 1.39(1.15to1.67) 0.001 1.65(1.16to2.34) 0.005

Higher 2.65(2.19to3.20) <0.001 2.74(1.94to3.86) <0.001

Durationofsymptoms(months)

<1 1.00(reference) 1.00(reference)

1to3 0.65(0.54to0.78) <0.001 0.70(0.50to0.98) 0.04

3to12 0.58(0.47to0.72) <0.001 0.49(0.31to0.78) 0.003

>12 0.57(0.45to0.73) <0.001 0.54(0.40to0.72) <0.001

Recurrentsymptoms:yes 1.24(1.04to1.47) 0.01 –

CI,confidenceinterval.

Tocheckwhethermissingdatabiasedtheresults,a regres-sionanalysiswasperformed,includingdummyvariablesfor missing data for each variable. The dummy variables for missingdataofurbanisationandeducationweresignificantly associatedwithself-referralinthe finalmodel for patients withankle complaints;missingurbanisation datawas sig-nificantlyassociatedwithself-referralinthefinalmodelfor patientswithkneecomplaints.However,thesefindingsdid notinfluencetheORsoftheothervariablesincludedinthe models.Thehighpercentageofmissingdataforeducation levelcanbeexplainedbythefactthatthisvariablewasonly createdinpatientsaged>16years.

Animportantlimitationofthisstudyisthattheseverityof complaintswasnotmeasured.As‘severityofsymptoms’has beensuggestedasapossibleexplanatoryfactorforthelower numberoftreatmentsessionsinpatientsthatself-referred,the interpretationof‘treatmentduration’and‘numberofvisits’ isquestionable[13,22,25].Inaddition,thevariables‘goals accomplished’and‘reasonsfor termination ofcare’ hada highpercentageofmissingdata(missing>50%).Duetothe highpercentage of missing data,these variableswere not includedinthemultivariateregressionanalyses.

Futureresearchshouldaimtoidentifydifferencesinthe reasonsforterminationofcare,anddifferencesinthe sever-ity of symptoms (e.g. pain and function scores) between patientswithkneeorankledisorderswhoself-referred com-paredwithpatientswhowerereferredbyaGP.Theseverity ofsymptoms(e.g.basedontheInternationalClassification

of Functioning,Disability andHealth)[26]andfunctional capacity[27,28]couldatleastbeassessedatthebeginning andendofatreatmentsession,butdiagnosisandgradingof aninjurywillalsoprovideessentialinformation.The prede-finedtreatmentgoalcouldbebasedonthemagnitudeofthe differenceinsymptoms beforeandafter thetreatment ses-sion.Bystandardisingthesetreatment goals,differencesin treatmentoutcomesbetweenpatients whoself-referredand thosewhowerereferredbyaGPcanbedetermined.

In countries whereself-referralis already implemented

[29],researchersshouldbeawareofthedifferencesinpatient characteristicsbetweenpatientswhovisitaphysiotherapist withorwithoutaGPreferral,asshowninthepresentstudy. These differencescould lead tothe selection of aspecific patientpopulationwhenrecruitinginasinglesetting.Finally, the results of the present study are important for physio-therapists andGPsas theyneed tobefully aware oftheir responsibilitiesfollowingtheimplementationofself-referral.

Conclusion

Since2006,patientshaveincreasinglyusedself-referral tovisitaphysiotherapist; self-referralrates stabilisedafter 2009.Theuseofself-referralamongpatientswithknee/ankle symptoms is associated with younger age, higher educa-tionlevelandashorterdurationofsymptoms.Furthermore, treatment characteristics (e.g. number of treatment

ses-sions, durationof treatment) differedbetweenGP-referred and self-referred patients. Future studies should examine theeffectivenessofphysiotherapyfollowingself-referralin patientswithacutekneeandanklesymptoms.

Ethicalapproval:Ethical approvalwasnotsoughtbecause NPCDdoesnotfallwithinthescopeoftheMedicalResearch InvolvingHumanSubjectsAct.

Funding:Thisstudywaspartlyfinanciallyendorsedbythe DutchArthritisFoundationascenterofexcellence “Optimiz-ingcareforosteoarthritisandotherchronicmusculoskeletal disorderinprimarycare”.

Conflictofinterest:Nonedeclared.

AppendixA. Supplementarydata

Supplementary data associated with this article can be found,inthe online version,athttp://dx.doi.org/10.1016/j. physio.2017.03.008.

References

[1]DavisIS,PowersCM.Patellofemoralpainsyndrome:proximal,distal, andlocalfactors,aninternationalretreat,April30–May2,2009,Fells Point,Baltimore,MD.JOrthopSportsPhysTher2010;40:A1–16. [2]SpronkI,UrsumJ,DavidsR,StirbuI,NielenMMJ,KorevaarJC,etal.

Huisarts–Top-20diagnosesbijcontactenmetdehuisartsenpraktijk naargeslacht.Uit:NIVELZorgregistratieseerstelijn;2013.Available at:www.nivel.nl/node/3444.[Accessed16May2017].

[3]VanderLindenMW,Westert GP,DeBakken DH,SchellevisFG. TweedeNationale Studienaarziektenenverrichtingeninde huis-artspraktijk. Klachten en aandoeningen in de bevolking en in de huisartspraktijk.Utrecht/Bilthoven:NIVEL/RIVM;2004.

[4]BeloJ, BuisP, van RijnRM,Sentrop-Snijders E, Steenhuisen S, WilkensC,etal.NHG-StandaardEnkelbandletsel(tweedeherziening). HuisartsWet2012;55(8).

[5]Belo JN, Berg HF, Klein Ikkink AJ, Wildervanck-Dekker CMJ, Smorenburg HAAJ, Draijer LW. NHG-Standaard Traumatische knieproblemen(Eersteherziening).HuisartsWet2010;54:147–58. [6]BeloJN,Bierma-ZeinstraSMA,RaaijmakersAJ,VanderWisselF,

OpsteltenW. NHG-StandaardNiet-traumatische knieproblemenbij volwassenen(Eersteherziening).HuisartsWet2008;51:229–40. [7]BreedveldtBoerHP,KlaassenWRC,SpinnewijnWEM,HeinenN,

BurggraaffHB,DerksCJT,etal.NHG-StandaardNiet-traumatische knieproblemenbijkinderenenadolescenten(Eersteherziening). Huis-artsWet2009;52:332–41.

[8]McAlindonTE,BannuruRR,SullivanMC,ArdenNK,Berenbaum F,Bierma-ZeinstraSM,etal.OARSIguidelinesforthenon-surgical managementofkneeosteoarthritis.OsteoarthrCartil2014;22:363–88. [9]vanLinschotenR,vanMiddelkoopM,BergerMY,HeintjesEM, Ver-haarJA,WillemsenSP,etal.Supervisedexercisetherapyversususual careforpatellofemoralpainsyndrome:anopenlabelrandomised con-trolledtrial.BMJ2009;339:b4074.

[10]vanRijnRM,vanOchtenJ,LuijsterburgPA,etal.Effectivenessof additionalsupervisedexercisescomparedwithconventionaltreatment aloneinpatientswithacutelateralanklesprains:systematicreview. BMJ2010;341:c5688.

[11]KooijmanMK,BartenJA,VerberneLDM,LeemrijseCJ,VeenhofC, SwinkelsICS.Fysiotherapie–wijzevantoegang.Utrecht:NIVEL; 2013.Availableat:www.nivel.nl/node/3238.[Accessed16May2017]. [12]BartenJA,Koppes LLJ.Zorg doorde fysiotherapeut – wijzevan toegang.Utrecht:Nivel;2017.Availableat:www.nivel.nl/node/4674. [Accessed16May2017].

[13]LeemrijseCJ,SwinkelsIC,VeenhofC.Directaccesstophysical ther-apyintheNetherlands:resultsfromthefirstyearincommunity-based physicaltherapy.PhysTher2008;88:936–46.

[14]ScheeleJ,VijfvinkelF,RigterM,SwinkelsIC,Bierman-ZeinstraSM, KoesBW,etal.Directaccesstophysicaltherapyforpatientswithlow backpainintheNetherlands:prevalenceandpredictors.PhysTher 2014;94:363–70.

[15]SwinkelsIC,KooijmanMK,SpreeuwenbergPM,BossenD, Leemri-jseCJ,DijkCE,etal.Anoverviewof5yearsofpatientself-referral forphysicaltherapyintheNetherlands.PhysTher2014;94(December (12)):1785–95.

[16]WOMACInternational.Internationalclassificationofprimarycare.2nd ed.Oxford:OxfordUniversityPress;2003.

[17]CimminoMA,FerroneC,CutoloM.Epidemiologyofchronic muscu-loskeletalpain.BestPractResClinRheumatol2011;25:173–83. [18]Hagen KB,Bjorndal A, Uhlig T, Kvien TK.A population study

of factors associated with general practitioner consultation for non-inflammatorymusculoskeletal pain.AnnRheumDis 2000;59: 788–93.

[19]JordanKP,KadamUT,HaywardR,PorcheretM,YoungC,CroftP. Annualconsultationprevalenceofregionalmusculoskeletalproblems inprimarycare:anobservationalstudy.BMCMusculoskeletDisord 2010;11:144.

[20]KasteleinM,LuijsterburgPA,BeloJN,VerhaarJA,KoesBW, Bierma-ZeinstraSM. Six-yearcourse andprognosis of nontraumaticknee symptomsinadultsingeneralpractice:aprospectivecohortstudy. ArthritisCareRes(Hoboken)2011;63:1287–94.

[21]O’ConnorSR,BleakleyCM,TullyMA,McDonoughSM. Predict-ingfunctionalrecoveryafteracuteanklesprain.PLoSOne2013;8: e72124.

[22]MitchellJM,deLissevoyG.Acomparisonofresourceuseandcost indirectaccessversusphysicianreferralepisodesofphysicaltherapy. PhysTher1997;77:10–8.

[23]OjhaHA,SnyderRS,DavenportTE.Directaccesscomparedwith referredphysicaltherapyepisodesofcare:asystematicreview.Phys Ther2014;94:14–30.

[24]HoldsworthLK,WebsterVS.Directaccesstophysiotherapyinprimary care:now?—andintothefuture?Physiotherapy2004;90:64–72. [25]PendergastJ,KliethermesSA,FreburgerJK,DuffyPA.Acomparison

ofhealthcareuseforphysician-referredandself-referredepisodesof outpatientphysicaltherapy.HealthServRes2012;47:633–54. [26]WorldHealthOrganization.Internationalclassificationoffunctioning,

disability,andhealth.Geneva:WHO;2001.

[27]deBieRA,deVetHC,vandenWildenbergFA,LenssenT,Knipschild PG.Theprognosisofanklesprains.IntJSportsMed1997;18:285–9. [28]TegnerY,LysholmJ.Ratingsystemsintheevaluationofkneeligament

injuries.ClinOrthopRelatRes1985;198:43–9.

[29]BuryTJ,StokesEK.Aglobalviewofdirectaccessandpatient self-referraltophysicaltherapy:implicationsfortheprofession.PhysTher 2013;93:449–59.

Availableonlineatwww.sciencedirect.com

ScienceDirect