Adherence
Adherence
to
and
beliefs
in
lipid-lowering
medical
treatments:
A
structural
equation
modeling
approach
including
the
necessity-concern
framework
Erik
Berglund
*
,
Per
Lytsy,
Ragnar
Westerling
DepartmentofPublicHealthandCaringSciences,UppsalaUniversity,Uppsala,Sweden
1. Introduction
Cardiovasculardisease(CVD)istheleadingcauseofdeathinthe industrializedworld[1,2].Dyslipidemiaisanimportantriskfactor forCVD,estimatedtocause18%ofcerebrovasculardiseaseand56% ofischemicheartdisease[3].Cholesterolloweringhasbeenthe primarygoaloftherapiesaimedatCVDriskreduction,andseveral randomized studies have demonstrated the benefits of statins (hydroxymethylglutaryl-CoA reductaseinhibitors) in the reduc-tion of cardiovascular-related events within high-risk patient groups [4].Currently, statindrugtreatment is one ofthe most importanttreatmentstrategieswhenmanagingpatientswith,or athighriskof,CVD.
Adherenceisdefinedastheextenttowhichaperson’sbehavior, suchastakingmedication,followingadietorexecutinglifestyle
changes, correspondswiththerecommendationsfrom a health care provider [5]. Poor adherence has been shown to be an importantfactorfortreatmentfailurewhenlookingatbothhigh cholesterollevels[6]andmorbidity[7–9],and,asaresult, non-adherencetotreatmentisconsideredtobeacardiovascularrisk factor[10].Adherencetolong-termpharmacologicaltherapyfor chronicillnessesindevelopedcountriesaverages50%[5],andfor lipid-lowering pharmacological therapies the long-term adher-enceispooranddecliningconsiderablyovertime.
In 2003, the World Health Organization (WHO) described adherence as a phenomenon determined by five dimensions: patient-related factors,social and economicfactors,health care team and system-related factors, condition-related factors and therapy-related factors [5]. To describe adherence and for the analysisofnon-adherenceamongpatientswithCVD,hypertension andotherlong-termtherapies,alargenumberofhypothesesand factorshavebeenproposed[11].
Severalmodelsthataimtoexplainhealthbehaviorarebasedon patients weighing positive and negative perceptions for a treatmentorhealthadvice,wherethebalancedirectsthebehavior. ThemodelsthatbeenusedinadherencestudiesaretheHealth
ARTICLE INFO Articlehistory: Received8March2012
Receivedinrevisedform29October2012 Accepted4November2012 Keywords: Cardiovasculardisease Dyslipidemia Prevention Lipid-lowering Statins Treatmentadherence Patientexpectations Necessity Concern
Healthlocusofcontrol Prescription Pathway Pathanalysis SEM PLS ABSTRACT
Objective:Thisstudyattemptstoidentifyastructureamongpatient-relatedfactorsthatcouldpredict
treatmentadherenceinstatinpatients,especiallywithregardstothenecessity-concernframework.
Methods:414Swedishpatientsusingstatinscompletedaquestionnaireabouttheirhealth,treatment,
locus of control, perception of necessity-concern and adherence. The data were handled using a
structuralequationmodelingapproach.
Results:Patientsthatreportedhighperceptionsofnecessitytotreatmentseemedtoadherewell,and
sideeffectsappeartoaffectadherencenegatively.Diseaseburden,cardiovasculardiseaseexperienceand
highlocusofcontrolseemtohavemediatingeffectsonadherence.
Conclusion:Thisstudyprovidessupportforthehypothesisthathealth-andtreatment-relatedfactors,as
wellaslocusofcontrolfactors,areindirectlyassociatedwithtreatmentadherenceviatheirassociation
withmediatingfactornecessity.
Practice implications:This study highlights the importance of considering patients’ beliefs about
medications,diseaseburden,experienceofcardiovasculareventsandlocusofcontrolasthesefactorsare
associatedwithadherencebehaviortostatintreatment.Thisstudyalsoemphasizesmoregenerallythe
importanceofanapproachtargetingnecessityandconcernwhencommunicatingwithandtreating
patientswithlipid-loweringmedication.
ß2012ElsevierIrelandLtd.
*Correspondingauthorat:DepartmentofPublicHealthandCaringSciences, UppsalaUniversity,Box564,SE-75122Uppsala,Sweden.Tel.:+46184716553; fax:+46184716675.
E-mailaddress:erik.berglund@pubcare.uu.se(E.Berglund).
ContentslistsavailableatSciVerseScienceDirect
Patient
Education
and
Counseling
j ou rna l hom e pa ge :ww w. e l s e v i e r. c om/ l o ca t e / pa t e duc ou
0738-3991ß2012ElsevierIrelandLtd.
http://dx.doi.org/10.1016/j.pec.2012.11.001
Open access under CC BY-NC-ND license.
Belief Model [12,13], the Transtheoretical Model [14], the Protection Motivation Theory [15,16] and the Self-Regulatory Model (SRM) [17,18]. The SRM proposes that health-related behaviors are cognitive responses influenced by a patient’s perceptionof treatment and emotional response to treatment. Theseresponsescanbederivedfrombothmanifestsymptomsand concernaboutahealththreat,orexperienceorconcernaboutside effectsfromatreatment.
Influenced by the earlier models, the necessity-concern framework(NCF)wasdevelopedtospecificallyinvestigatedrug treatment adherence [19]. According to the NCF, a patient’s decisionregardingadherenceistheresultofatrade-offbetween thepatient’sperceivedneedforaprescribedtreatment(necessity) andtheirworriesaboutthepotentialadverseeffectsasaresult (concern).Inthisstudy,wechosetoassesspatients’beliefsusing the NCF as it has been used in a broad range of different quantitativestudiesexploringdrugtreatmentadherence[20–23], especiallyforcardiovasculardiseases[24–27].
Somefactorsseemtobemorerelatedthanothers.Factorswith a high probability of affecting adherence include gender [28], demographics[29,30],patient understanding and perceptionof medication[5],sickness-and treatment-relatedfactors[31–34], andhealthlocusofcontrol[35].Thehealthlocusofcontrolmodel isdefinedbythreedifferentdimensions:anindividual’ssenseof controlovertheirhealthisdirectlyrelatedtotheirownbeliefsand actions (internal); to chance externality (chance); or to the influenceofotherimportantpersons(powerfulothers)[36].There issupportfortheideathataperson’slocusofcontrolisassociated withhealthbehavior,mainlyincombinationwithotherpredictive factors[37]. Qualitative studiessuggest that individualswitha stronglocusinpowerfulothers mightbemoreadherent tothe recommendationsofhealthcareprofessionals[38].
Todate,howthese(andother)factorsarerelatedtoadherence and non-adherence for patients with CVD has not been fully explored,andthereislittleinformationavailableregardinghow strongtheinfluenceofthesefactorsisonadherenceinadjusted models.Thisstudyattemptstoidentifyastructureamongfactors regarding demographic, health and treatment factors, locus of control,NCFandadherenceinpatientsusingstatins.Theaimisto present a model that describes the relationships between the central variables and a measurement structure that possibly predictsadherencewithinpatientgroupsathighriskofCVD. 2. Methods
For this study,across-sectional studydesignwasapplied.A totalof600postalquestionnairesweredistributedinMay2009to the 28 operating pharmacies within the county of Uppsala in centralSweden.Thenumberofquestionnairesdistributedtoeach pharmacy was proportional to the number of previous statin prescription sales. The employees of each pharmacy were instructedtoinviteeverypatientwho visitedthepharmacyfor the preparation of their statin prescription. There were no inclusion criteria other than the statin prescription requisite, andno exclusion criteria. Patientsagreeing toparticipate,after receiving oral and written information about the studyby the pharmacist, were handed a questionnaire to take home and complete, and then return by post. The number of patients declining to participate was registered for control of non-participants.Thefirstpageofthequestionnairecontainedprecise information on the purpose of the study.Completed question-naires were returned anonymously in a prepaid envelope. All questionnairesreturnedwithinthreemonthswereincludedinthe study.Atotalof697statinuserswereaskedtoparticipate:109 declined to participate and 588 questionnaires were handed out (one pharmacy failed to distribute their questionnaires).
Questionnaires were returned by 414 individuals, making the responserateofthedistributedquestionnaires 70.4% (414/588) andtheoverallresponserate59.4%(414/697).
2.1. Measuresinquestionnaires
Thequestionnairecontainedatotalof76questions.Themain datatypesandmeasuresthatwereincludedwere:
Demographic data: This was collected using questions that assessedtherespondent’sgender,age,occupationandeducational level,includingcompulsoryschool,secondaryschool(or equiva-lent)anduniversity.
Health-, disease- and treatment-related factors: Data were collectedusinga listof14 commonhealthproblems(usedasa cumulative measure of disease burden and number of health problems),cardiovasculardiseaseexperience(myocardial infarc-tion and/or angina), perceived satisfaction with treatment explanationsmadebyaphysician,andtimeonstatintreatment; thesequestionshavebeenusedearlier[39].Experiencesorworries ofside effectsand difficultiesswallowingsolid dosescanaffect adherencenegatively[34],anddatawereassessedbythequestion: Doyouexperienceanyofthisunpleasantnesswhentakingyour statins?(a)Yes,IfeelthatIhavetroubleswallowingtablets,(b) Yes,IfeelthatIencounterunpleasantsideeffectsfromthem,(c) No,Idonotfeelanyunpleasantreactionsrelatedtomytreatment. Thevariablewasscoredasacountvariable.
Health locusof control:Thesedata weremeasured usingthe Multidimensional HealthLocus of Control (MHLC) 18-itemtest [36].MHLCisameasurementinstrumentthatincludesthree six-pointLikertscales:Internal(MHLC-I),Chanceexternality (MHLC-C)andPowerfulothers(MHLC-PO).Thedifferentscales,orlevels, wereanalyzed separately. In this study, theMHLC scaleswere treatedasindexonlyinthecorrelationmatrix.
Beliefsaboutmedicines:ResultsweremeasuredusingNCFbased on theBeliefs about Medicines Questionnaire-Specific (BMQ-S) [19].BMQ-Sisavalidated10-itemtestinstrumentwhichassesses beliefs about perceived medication necessity and perceived medication concernson five-point Likert scales.BMQis a two-scaleconstructionandisalsoavailabletouseasanindex.Inthis study,theindexwasonlyusedinthecorrelationmatrix.TheBMQ questionnaire has been translated into Swedish, with a back translationapprovedbytheoriginalauthorofthequestionnaire, andhasbeenpreviouslyusedinSweden[40–43].
Medicationadherence:Thesedatawereself-reportedusingthe Moriskyscaleofadherence(MSA) inafour-itemform[44].The MSAis a countvariableand thefirstquestionis:‘‘Doyou ever forgettotakeyourmedicine?’’.TheMoriskyscalewasoriginally designed to evaluate medication adherence in hypertensive patients, but has subsequently been found to be reliable in a varietyofadherencestudies[45,46].Inpreviousstatinstudies,the MSAusedwasbinary,withonlytwocategories[47].Patientswho answered ‘‘no’’ to all questions were categorized as highly adherent, while patients who answered ‘‘yes’’ to at least one question were categorized as having low adherence. This categorization system is consistent withwhat was used when developingtheoriginalscale,aswellashowithasbeenusedin severaladherencestudies[47,48].
2.2. Methodofdataanalysis
The Statistical Package for the Social Sciences version 19 (Chicago, IL, USA) was used for descriptive statistics, factor analysis,tomeasurethevarianceinflationfactor(VIF),and Chi-squareandMann–WhitneyUtests.WarpPLSvs.2.0wasusedfor structuralequationmodeling(SEM)analysiswiththepartialleast squares(PLS)estimationtechnique[49].SEMisacombinationof
confirmatoryfactorsandpathanalysis,whichallowstheinclusion oflatentvariables(LV)thatarenotdirectlymeasured[50].SEM workswithboth continuousand discreteobserved variablesas indicators (LVs). SEMis a second-generation statistical method that, in contrast to regression, allows for the simultaneous assessment of multipleindependent and dependent constructs, includingmulti-step paths[51]and mediatingeffects [52].LVs differfromtheobservedsumscores(index)oftheindicatorsas theycanaccountformeasurementerrorsintheitems,anditems arealloweddifferentialweightsinestimatingthelatentconstruct [53].Inessence,LVscanbeformativeorreflective.Thedifferenceis
in the directionof theoreticalcausality between measures and constructs. Reflectivemeasures are theoretically caused by the latentconstruct,whereasformativemeasurestheoreticallycause thelatentconstruct[54].
SEMwasconductedusingthePLSestimationtechniquewith Wold’s algorithm [55–57]. PLS is a modeling approach witha flexible technique, which canhandle datawith missing values, stronglycorrelatedvariablesandsmallsamples.SEM-PLSisa well-suitedmethodforexploratoryresearchandtheorydevelopment [58],whichwasthepurposeofthisstudy.SEM-PLShasalsobeen usedforadherencestudies[59,60].SEMworkswithtwomodels:
Table1
Characteristicsofthestudygroup(n=414).
Highadherent Lowadherent Total P
Sex Male(%) 27.6 23.2 50.8 0.898a
Female(%) 27.0 22.2 49.2
Age,mean(s.d.) 64.6(10.0) 63.3(9.1) 64.2(9.5) 0.197b
Education Compulsoryschool(%) 21.9 18.0 39.9 0.636a
Secondaryschool(%) 16.7 12.3 29.1 University(%) 16.0 15.0 31.0
Occupation Full-orpart-timework(%) 21.5 19.6 58.9 0.391a
Notinworkforce(%) 33.4 25.5 41.1
Diseaseburden Low(%) 15.9 17.0 32.9 0.234a
Medium(2–4)(%) 27.3 19.7 47.1
High(%) 11.1 8.9 20.0
Cardiovasculardisease Noexperience(%) 37.9 34.6 72.6 0.184a
Experience(%) 16.4 11.0 27.4
Satisfactionwithtreatmentexplanation Noneorpoor(%) 4.2 4.4 8.6 0.506a
Fair(%) 9.8 9.6 19.4
Goodorverygood(%) 40.5 31.4 72.0
Treatmenttime Lessthan2years(%) 20.4 17.9 38.3 0.850a
Between2and5years(%) 14.7 12.5 27.3 Morethan5years(%) 19.4 15.0 34.4
Sideeffects No(%) 50.4 37.7 88.0 0.021a Yes(%) 4.7 7.2 12.0 MHLC,indexform Internal(MD) 23 23 23 0.513b Chance(MD) 17 16 17 0.820b Powerfulothers(MD) 20 19 19 0.050b BMQ,indexform Necessity(MD) 17 15 17 0.000b Concerns(MD) 11 11 16 0.133b a Chi-squaredtest. b Mann–WhitneyUtest. Necessity-Concern Framework Internality Chance Powerful others
NECESSITY CONCERN
Muldimensional Health Locus of Control Demographic
Gender Age Educaon Occupaon
Health- and treatment-related factors Disease burden CVD experience Explanaon sasfacon Treatment me Sideeffects
Behaviorrelated to Medicaon
ADHERENCE Background variables Mediang variables Dependent variable
Fig.1.Researchframework.
ThisBATLoCmodeloutlinesthetheoreticaldirectandindirectrelationshipsbetweenthebackgroundvariables:demographics,health-andtreatment-relatedfactors,MHLC andthemediatingvariablesinNCFandthedependentvariableadherence.
(I)ameasurementmodel(alsocalledthe‘‘outermodel’’),which determines the relationships between observed manifesting variables and their association withlatentvariables; and (II) a structuralmodel(also calledthe‘‘inner model’’),relating latent variablestootherlatentvariables.PLSestimatesloadingandpath parametersbetweenlatentvariablesandmaximizesthevariance explainedforthedependentvariables.TheWarpPLSprogramcan handlelinearaswellasS-and U-shapedrelationships between variables.Thepathsinthemodelweretestedforsignificanceusing the bootstrapping procedure, with 200 cases of resampling
incorporated in WarpPLS. Significant mediating effects were calculatedusingtheSobeltest[61].
Model fit indicators are important in SEM since they offer comparablemeasurements.Modelfitindicatorsapplytothedegree of correspondence between the observed data and the model-implieddata.The degreeof correspondenceis determined bya functionofthesumofthesquareddeviationsbetweentheobserved samplecovariancematrixandthemodel-impliedcovariancematrix. In WarpPLS, the output model fit is assessedby three indices: average path coefficient (APC), average R-squared (ARS) and
Fig.2.SEManalysisofdataoutlinedaftertheframework. (*)Patchissignificantatthe0.05level.
(**)Patchissignificantatthe0.01level. (***)Patchissignificantatthe0.001level.
SEManalysisgeneratedthroughapartialleastsquaresestimationtechnique,withpathcoefficientsofthestructuralpathwaymodel(i.e.innermodel).Themodeloutlinesthe hypnotizedrelationshipsamongthelatentvariablesintheBATLoCmodel.Totestwhetherdirecteffectshadanimpactonadherence,thedemographicfactors,MHLCand health-anddisease-relatedfactorsweretesteddirectlyagainstadherenceinthemodel.Significantassociationsbetweenlatentvariablesarepresentedinbold. Table2
Correlationanalysisamongindicators(observedvariables).
Min Max Std.dev. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1.Gender 1 2 .50 2.Age 22 89 9.52 .06 3.Educationlevel 1 3 .84 .09 .27** 4.Occupation 0 1 .49 0 .73** .25** 5.Diseaseburden 1 3 .72 .05 .06 .02 .13* 6.CVDexperience 0 1 .45 .16** .18** .12* .11* .18**
7.Treatmentexplanationsatisfaction 1 5 .97 .06 .04 .13**
.09 .04 .01 8.Treatmenttime 1 3 .85 .09 .24** .08 .22** .19** .24** .01 9.Sideeffects 0 1 .32 .02 .06 .11* .10* 0.02 .02 .02 .00 10.MHLCInternal 6 36 5.37 .17** .05 .02 .03 .13* .10* .09 .06 .04 11.MHLCChance 6 33 4.59 .03 .12* .27** .15** .12* .01 .06 .11* .02 .18** 12.MHLCPowerfulother 7 36 5.26 .22** .24** .11* .17** .02 0.10 .14** .11* .02 .22** .21** 13.Necessity 5 25 3.82 .01 .11* .13** .12* .26** .20** .07 .25** .00 .02 .10* .27** 14.Concern 5 23 4.30 0 .09 .04 .08 0.08 .07 .20** .01 .18** .02 .10* .08 .23** 15.Adherence 0 1 .50 .01 .06 .02 .04 0.05 .07 .07 .02 .12* .03 .01 .10 .22** .08
Max.min.standarddeviationsandcorrelationsindicators.ThismatrixhasbeencalculatedwithSpearman’ssignificanttest.Indiceswereusedformultidimensionalhealth locusofcontrol,perceptionofnecessityandconcernandadherence.Allindicatorsinthemodelwerecodedsothat(hypnotized)higherscoresontheindependentvariable alsoindicatehigherscoresontheoutcomevariable,iftherewasapositiveassociation.
*
Correlationissignificantatthe0.05level.
**
average variance inflation factor (AVIF). The main reason why WarpPLS includes APC and ARS is to enable an acceptable comparisonbetweendifferentmodels,whichiswhythesemeasures areoflowerimportanceinstudies likethis,whereeachpathis independentlyimportant.However,figuresforAPCandARSshould both be under 2 and should both be statistically significant (p<0.05),whilethevalueforAVIFisrecommendedtobebelow5. 2.3. Researchframeworkandmodelconstruction
A research model of balanced adherence influenced by treatmentandlocusofcontrolfactors(BATLoC)wasconstructed toexaminetherelationships betweenthevariables (Fig.1).The model contains one dependent (adherence), two mediating (perceptionofnecessityandperceptionofconcern),and twelve independentLVs.ThetwelvebackgroundLVsweredividedinto threegroups:demographicvariables(gender,age,educationlevel andoccupation);health-andtreatment-relatedvariables(disease burden,cardiovasculardiseaseexperience,treatmentexplanation satisfaction,treatmenttimeandsideeffects);andhealthlocusof controlvariables(onthreelevels:internal,chanceandpowerful others).
3. Results
The average age of the study population was 64.2 years (S.D.9.5),andthegroupconsistedofslightlymoremen(51.1%) thanwomen(48.9%).Compulsoryschool wasthemostcommonly completededucationlevel(40.0%).Approximately40.6%ofthegroup wereinfull-timeorpart-timework,whiletheremaining59.4%were unemployedorretiredfromthe workmarket. Thedistribution of demographicsandkeyvariablesinthestudypopulationisshownin Table1.
3.1. Adherenceandcardiovasculardiseases
Inthewholegroup,54.5%ofpatientswereclassifiedtohave highadherence,and45.5%wereclassifiedtohavelowadherenceto theirstatintreatment.Aboutone-fifthofthegroupreportedahigh diseaseburden(sufferingfromfiveormorediseases)andhalfof thegrouphadbetweentwoandfourdiseases.Overall,72.8%ofthe patientsdidnotreportanyCVDexperience,andthereforereceived theirtreatmentasprimaryprevention,27.2%ofthegroupreported at least one CVD experience, so received their treatment as secondaryprevention.Themajorityofthegroupdidnotreportany sideeffects,but11.9%didexperiencesomesideeffects.
3.2. Multidimensionalhealthlocusofcontrolandbeliefsabout medicines
The Mann–WhitneyUtest in Table1 showedno significant differenceon internalorchancebetweenpatientswithlow and highadherence,only smalldifferenceswereseen ontheMHLC indexscales.
3.3. Correlationmatrix
Severaloftheassociationsoutlinedintheresearchframework (Fig.1)werealsosignificantinthecorrelationmatrix(Table2).The highestcorrelationtotheadherencevariableswasseenwiththe perceptionofnecessityoftreatment.Theindicatorvariableswere testedformulticollinearity,andnovariablehadover2.5in VIF, which indicates that the risk for multicollinearity can be considered to be low. These imply acceptability of using a structuralequationmodel.
3.4. FullmodelandPLS-SEManalysis
A PLS estimation procedure was used to examine the hypothesizedrelationships (Fig.2)between constructsdepicted inthetheoreticalframework(Fig.1).TheSEManalysisshoweda significant relationship between adherence and necessity of treatment (
b
=0.15, p=0.010),but notwith concern (Table 3). The explanatory variables were also tested directly against adherence, and it was found that side effects (b
= 0.14, p=0.006)hadasignificanteffectonadherence.Theanalysisshowedthateducationlevel(
b
= 0.10,p=0.033), disease burden (b
=0.20, p<0.001), CVD experience (b
=0.17, p<0.001),satisfaction with treatment explanations made bya physician (b
=0.13, p=0.008), treatment time (b
=0.14, p<0.001) and powerful others in locus of control (b
=0.33, p<0.001)each had aneffect on perceptionofthenecessity of treatment. In addition, education level (b
= 0.09, p=0.017), satisfaction with treatment explanations made by a physician (b
= 0.26, p<0.001),sideeffects(b
=0.17,p<0.001),MHLC-C (b
=0.09,p=0.025)andMHLC-PO(b
=0.14,p=0.001)allhadan effectonconcern.Intotal,thesevariablescouldexplainalmost31% ofthevarianceseeninperceptionofnecessity(R2=0.31)and16% ofthevarianceseeninperceptionofconcern(R2=0.16)and6%ofthevarianceseeninadherence(R2=0.06).
Three background LVs had significant mediating effects on adherence (through necessity of treatment): disease burden (
b
=0.03, p=0.034), CVD experience (b
=0.03, p=0.034) and powerfulothers(b
=0.05,p=0.019).Table3
PathcoefficientsandP-valuesofdirecteffectsonNCFandadherence.
Backgroundvariables Mediatingvariables Dependentvariable
Necessity Concern Adherence
Pathcoefficients P-Value Pathcoefficients P-Value Pathcoefficients P-Value
Gender 0.04 0.199 0.06 0.124 0.02 0.310 Age 0.02 0.188 0.06 0.356 0.04 0.269 Educationlevel 0.10 0.033 0.09 0.017 0.02 0.358 Occupation 0.03 0.486 0.01 0.303 0.01 0.407 Diseaseburden 0.20 <0.001 0.06 0.328 0.02 0.378 CVDexperience 0.17 <0.001 0.06 0.108 0.02 0.242 Explanationsatisfaction 0.13 0.008 0.26 <0.001 0.06 0.127 Treatmenttime 0.14 <0.001 0.08 0.272 0.05 0.225 Sideeffects 0.04 0.102 0.17 <0.001 0.14 0.006 MHLCinternal 0.00 0.363 0.03 0.325 0.05 0.179 MHLCchance 0.02 0.416 0.09 0.025 0.05 0.191 MHLCpowerfulother 0.33 <0.001 0.14 0.001 0.05 0.191 Necessity 0.15 0.010 Concern 0.08 0.067
3.5. Testofpathmodel
Thewholemodeldemonstratedanacceptablefittothedatafor APC=0.081(p<0.001),ARS=0.176(p<0.001)andAVIF=1.269.
4. Discussionandconclusion
This studyaimedtocreateand examinea modelthat could contributetotheunderstanding andpredictabilityofadherence withinpatientgroupsatriskofCVD.Anewmodelandstructure wasoutlinedthattestedtheassociationsofdemographics,health andtreatment,locusofcontrolonNCFandadherence.Mostfactors includedwerealreadyknowntohaveanimpactonadherence.A primaryaimwastocreateamodelthatcouldhandlethewhole frameworksimultaneously.
4.1. Discussion
Inthisstudyofstatinusers,highbeliefintreatmentnecessity hasapositiveassociationwithadherence,whileconcernsabout treatmentseemtohavelittleassociationwithadherentbehavior. Thisindicatesthat patientsseem toattachmore importanceto factorsotherthananegativeassociationwithdrugswhenitcomes toactualtreatmentbehaviors.
Amongthebackgroundvariables,diseaseburden,CVD experi-ence,treatmenttimeandpowerfulothersinlocusofcontrolseem tohavepositiverelationshipswithbeliefintreatmentnecessity. Threeofthebackgroundvariablesalsohadasignificantmediating effectonadherencethroughtheperceptionofnecessity:disease burden,CVDexperience and locus ofcontrol throughpowerful others.Thismeansthatthesefactorshavea positiveimpacton adherence behavior through mediating necessity of treatment. Thesefindingsareinterestinginseveralways,especiallyasfactors that increase sickness severity seem to increase the perceived necessity of treatment, and therefore contribute to a higher adherence.Thisislogical,sinceapatientathigherriskofadisease alsohasmoretogainfromarisk-loweringtreatment.However,in earlierstudiesthisassociationdidnotbecomeevidentatapatient level[39].
Highereducationandsatisfactionwithtreatmentexplanations madebyaphysicianwerenegativelyassociatedwithconcernsthat thepatients heldabouttheirmedications.Sideeffectsand high beliefinchanceandpowerfulothersseemtoincreasetheconcerns that the patients reported about their medical treatment. Side effects and fear of potential side effects are well known tobe importantfactorsfornon-adherence[62].
A high satisfaction with the treatment explanation was associated with a higher perception of necessity of treatment andlowerconcernsabouttreatment.Thisisconsistentwithearlier studies which have shown that the communication of related issues between patients and physicians has an impact on adherence[30].Physiciansandhealthcarepersonnelingeneral mightalsobeviewedaspowerfulothersbypatients,whichisalso measured as a locus of control variable. Powerfulothers were positively associated with necessity of and concerns about treatment, with necessity showing the strongest association. These results are consistent with the study performed by GillibrandandFlynn,whofoundanassociationbetweenpowerful othersandtheabilitytocopewithlong-termtreatments[38].
Theresultsshowthatdiseaseburdenhadapositiveassociation withnecessityoftreatment,andamediatingeffectonadherence. Anexplanationcouldbethata personwithmany diseases has morecontact with health care providers,and is provided with moreinformationand encouragementin ordertomanage their healthcareproblems.
The factorsin this model explained6% ofadherence in this study.Thatmayseemlow,butitisinthesamerangethatother studieshaveshownforpatientsinthismedicalgroup[63].NCFhas ahigherpotential,andHorneandWeinmanindicatedthatpatient beliefs aboutmedications contributed toabout one-fifthof the totalvarianceintheadherencebehaviorofpatientswithchronic physical illness [32]. However, this indicates that adherence is associatedwithothervariablestoalargeextent.Anothertypeof adherencemeasurecouldpossiblyhaveobtainedadifferentresult, butadherenceisgenerallyacomplexbehaviortomeasure[64].
Four of the factors had more than one significant path. Experiences of side effects appeared to both lower adherence andincreaseconcern,andthisoutcomeseemslogical.Experience ofsideeffectswasalsotheonlybackgroundvariablethathada directimpactonadherence,whichisabehaviorthathasbeenseen inotherpatientgroupsaswell[65].Inaddition,satisfactionwith theexplanationoftreatmentsalsohadalogicalrelationshipwith theperceptionofnecessityandconcern,asitexplainednecessity andloweredconcerns.Educationallevelisnegativelyassociated withbothnecessityandconcerntoalmostthesamedegree,which shouldexcludetheeffectof thisvariable ina clinicalsituation. Indeeditdidnotappeartohaveanydirecteffectonadherence. Beliefinpowerfulothersshowedaninconsistentassociationwith necessityandconcern,asitincreasedboth,butnottothesame extent.Anexplanationforthiscouldbethatapersonwhohasgreat impressionsoftheirsurroundingsmightgetaccurateinformation regardingbothrisksandbenefits,whichincreasesnecessityand concern.
4.2. Limitations
Thisstudy wasofa cross-sectional type,which restrictsthe possibility of causal conclusions. The data on adherence to medication and NCF wereself-reported, and thereforesome of therespondentsmayhaveunderestimatedoroverestimatedtheir rateofadherence.
Theresearchmodelwasexplorative,andinfuturestudiesthe model may be complemented by other factors of interest, e.g. health beliefs [66,67], self-efficacy [68–72] and socioeconomic status[73],ortestedinothertheoreticalapproachestoinvestigate factorsofinterest.
This was a sample with limited diversity based on self-selection.Nodataonnon-respondentswerecollected.Tolimitthe impact of possible selection bias the model was adjusted for demographicvariablessuchasageandgender.Assuch,utilityand effectivenessamongdiversepopulationsshouldbeevaluatedin futureresearch.Inaddition,thispatientgroupwasselectedwhilst fetchingtheirprescribedmedications.Therefore,theresultsonly apply to secondary adherence behavior and should not be generalized to patients that are not primary adherent, which includes those patients who did not even purchase their prescriptiondrugs[74].
4.3. Conclusions
In conclusion, this study identified both the perception of necessity of treatment and side effects as directly significant factors associated with adherence among patients using lipid-loweringmedicaltreatments.Thisstudyalsoprovidedpreliminary supportforthenotionthathealth-andtreatment-relatedfactors, aswellaslocusofcontrolfactors,areindirectlyassociatedwith medical adherence through their associations with mediating perceptionofnecessityoftreatment.
Even though much of the adherence behavior is under the patients’ control [64], this result shows that perception of the necessity of treatment is associated with several modifiable
factors,andthatahighperceptionofthenecessityoftreatmentis associated with higher adherence among statin users. This supports the idea that present health care professionals have notseized thepotentialofincreasingadherence in this patient grouptoitsfullextent.
4.4. Practiceimplications
The study implies that it might be possible to increase adherencebymanagingsomeofthemodifiablefactorsthatare associatedwithCVDpatients’beliefsaboutmedications. Impor-tantly,patients’satisfactionwithtreatmentexplanationseemsto havea positiveassociationwithtreatmentnecessityandat the sametimeanegativeassociationwithtreatmentconcerns.
The studyhighlights theimportance for health care profes-sionalsofconsideringbeliefsaboutmedications,diseaseburden, experienceofcardiovascular eventsand locusof controlfactors that characterize the patient when it comes to increasing adherence. The results of this study imply that an approach targeting necessity and concern might be able to increase adherencetostatintherapy.
Conflictofinterest
Noneoftheauthorshaveaconflictofinteresttodeclare. Funding
Therewasnoexternalfundingforthisstudy. Ethicalapproval
EthicalapprovalwassoughtattheregionalEthicalCommittee ofClinicalInvestigationinUppsalabutwasnotdeemednecessary since the study group responded anonymously, leaving no possibilityofindividualidentification.
Acknowledgments
WewouldliketothanktheNationalCorporationofSwedish Pharmaciesand thestaffatthepharmaciesinUppsalafortheir assistance withthe distributionof questionnaires. We are also gratefultoRobertHorneforgrantingpermissiontousetheBMQ measurement, as well as to all of the respondents for sharing personalviewsabouttheirhealthandtheirtreatment.
References
[1]MichaudCM,MurrayCJ,BloomBR.Burdenofdisease—implicationsforfuture research.JAmerMediAssoc2001;285:535–9.
[2]RemmeWJ,SwedbergK.Guidelinesforthediagnosisandtreatmentofchronic heartfailure.EurHeartJ2001;22:1527–60.
[3]GuilbertJJ.Theworldhealthreport2002—reducingrisks,promotinghealthy life.EducHealth2003;16:230.
[4]BaigentC,KeechA,KearneyPM,BlackwellL,BuckG,PollicinoC,etal.Efficacy andsafetyofcholesterol-loweringtreatment:prospectivemeta-analysisof datafrom 90,056 participantsin 14randomised trials ofstatins. Lancet 2005;366:1267–78.
[5]WHO.Adherencetolong-termtherapies:evidenceforaction.Geneva:World HealthOrganization;2003.
[6]MaenpaaH,HeinonenOP,ManninenV.Medicationcomplianceandserum lipidchangesintheHelsinkiHeartStudy.BrJClinPharmacol1991;32:409–15. [7]BlackburnDF,DobsonRT,BlackburnJL,WilsonTW.Cardiovascularmorbidity associated with nonadherence to statin therapy. Pharmacotherapy 2005;25:1035–43.
[8]ChengCW,WooKS,Chan JC,Tomlinson B,YouJH.Association between adherencetostatintherapyandlipidcontrolinHongKongChinesepatients athighriskofcoronaryheartdisease.BrJClinPharmacol2004;58:528–35. [9]WeiL,WangJ,ThompsonP,WongS,StruthersAD,MacDonaldTM.Adherence
tostatintreatmentandreadmissionofpatientsaftermyocardialinfarction:a sixyearfollowupstudy.Heart2002;88:229–33.
[10]MungerMA,VanTassellBW,LaFleurJ.Medicationnonadherence:an unrec-ognizedcardiovascularriskfactor.MedGenMed2007;9:58.
[11]MyersL,MidenceK.Adherencetotreatmentinmedicalconditions.London: HarwoodAcademic;1998.
[12]RosenstockIM,StrecherVJ,BeckerMH.SociallearningtheoryandtheHealth BeliefModel.HealthEducQ1988;15:175–83.
[13]OlsenS,SmithS,OeiT,DouglasJ.Healthbeliefmodelpredictsadherenceto CPAPbeforeexperiencewithCPAP.EurRespirJ2008;32:710–7.
[14]Prochaska JO,Velicer WF.Thetranstheoreticalmodelofhealth behavior change.AmJHealthPromot1997;12:38–48.
[15]Palardy N,GreeningL, OttJ,Holderby A,AtchisonJ. Adolescents’health attitudesandadherencetotreatmentforinsulin-dependentdiabetesmellitus. JDevBehavPediatr1998;19:31–7.
[16]FloydDL,Prentice-DunnS,RogersRW.Ameta-analysisofresearchon pro-tectionmotivationtheory.JApplSocPsychol2000;30:407–29.
[17]LeventhalH,NerenzD,SteeleD.Illnessreperceptionsandcopingwithhealth threats.In:BaumA,TaylorSE,SingerJE,editors.HandbookofPsychologyand Health.Erlbaum:Hillsdale;1984.p.219–52.
[18]LeventhalH,DiefenbachM,LeventhalEA.Illnesscognition:usingcommon sensetounderstandtreatmentadherenceandaffectcognitioninteractions. CognitiveTherRes1992;16:143–63.
[19]HorneR,WeinmanJ,HankinsM.Thebeliefsaboutmedicinesquestionnaire: thedevelopmentandevaluationofanewmethodforassessingthecognitive representationofmedication.PsycholHealth1999;14:1–24.
[20]AikensJE,NeaseJrDE,NauDP,KlinkmanMS,SchwenkTL.Adherenceto maintenance-phaseantidepressantmedicationasafunctionofpatientbeliefs aboutmedication.AnnFamMed2005;3:23–30.
[21]HorneR,WeinmanJ.Self-regulationandself-managementinasthma: explor-ingtheroleofillnessperceptionsandtreatmentbeliefsinexplaining non-adherencetopreventermedication.PsycholHealth2002;17:17–32. [22]Llewellyn CD, MinersAH, Lee CA,Harrington C,Weinman J. Theillness
perceptionsandtreatmentbeliefsofindividualswithsevere haemophilia andtheirroleinadherence tohometreatment.PsycholHealth2003;18: 185–200.
[23]NeameR,HammondA.Beliefsaboutmedications:aquestionnairesurveyof peoplewithrheumatoidarthritis.Rheumatology2005;44:762–7.
[24]AllenLaPointeNM,OuFS,CalvertSB,MelloniC,StaffordJA,HardingT, et al. Association between patient beliefs and medication adherence following hospitalization for acute coronary syndrome. Am Heart J 2011;161:855–63.
[25]AllenLaPointeNM,OuFS,CalvertSB,MelloniC,StaffordJA,HardingT,etal. Changesinbeliefsaboutmedicationsduringlong-termcareforischemicheart disease.AmHeartJ2010;159:561–9.
[26]MagadzaC,RadloffSE,SrinivasSC.Theeffectofaneducationalinterventionon patients’knowledgeabouthypertension,beliefsaboutmedicines,and adher-ence.ResSocialAdmPharm2009;5:363–75.
[27]SudA,Kline-RogersEM,EagleKA,FangJ,ArmstrongDF,RangarajanK,etal. Adherencetomedicationsbypatientsafteracutecoronarysyndromes.Ann Pharmacother2005;39:1792–7.
[28]TurnerBJ,HollenbeakCS,WeinerM,TangSS.Aretrospectivecohortstudyof thepotencyoflipid-loweringtherapyandrace-genderdifferencesinLDL cholesterolcontrol.BMCCardiovascDisord2011;11:58.
[29]LowryKP,DudleyTK,OddoneEZ,BosworthHB.Intentionalandunintentional nonadherence to antihypertensive medication. Ann Pharmacother 2005;39:1198–203.
[30]YiannakopoulouE,PapadopulosJS,CokkinosDV,MountokalakisTD. Adher-encetoantihypertensivetreatment:acriticalfactorforbloodpressurecontrol. EurJCardiovascPrevRehab2005;12:243–9.
[31]LeventhalH,CameronL.Behavioraltheoriesandtheproblemofcompliance. PatientEducCouns1987;10:117–38.
[32]HorneR,WeinmanJ.Patients’beliefsaboutprescribedmedicinesandtheir roleinadherencetotreatmentinchronicphysicalillness.JPsychosomRes 1999;47:555–67.
[33]GeorgeJ,KongDC,ThomanR,StewartK.Factorsassociatedwithmedication nonadherenceinpatientswithCOPD.Chest2005;128:3198–204.
[34]Tordoff JM,BaggeML,GrayAR,CampbellAJ,Norris PT.Medicine-taking practicesincommunity-dwellingpeopleaged>or=75yearsinNewZealand. AgeAgeing2010;39:574–80.
[35]VoilsCI,SteffensDC,FlintEP,BosworthHB.Socialsupportandlocusofcontrol aspredictorsofadherencetoantidepressantmedicationinanelderly popu-lation.AmerJGeriatricPsychiatry2005;13:157–65.
[36]WallstonKA,WallstonBS,DeVellisR.DevelopmentoftheMultidimensional HealthLocusofControl(MHLC)Scales.HealthEducMonogr1978;6:160–70. [37]OberleK.Adecadeofresearchinlocusofcontrol:whathavewelearned?JAdv
Nurs1991;16:800–6.
[38]GillibrandW, FlynnM.Forcedexternalizationof controlinpeople with diabetes:aqualitativeexploratorystudy.JAdvNurs2001;34:501–10. [39]LytsyP,WesterlingR.Patientexpectationsonlipid-loweringdrugs.Patient
EducCouns2007;67:143–50.
[40]JorgensenTM,AnderssonKA,MardbyAC.Beliefsaboutmedicinesamong Swedishpharmacyemployees.PharmWorldSci2006;28:233–8.
[41]MardbyAC,AkerlindI,JorgensenT.Beliefsaboutmedicinesandself-reported adherenceamongpharmacyclients.PatientEducCouns2007;69:158–64. [42]MardbyAC,AkerlindI,HedenrudT.Differentdevelopmentofgeneralbeliefs
about medicinesduring undergraduatestudiesinmedicine,nursing and pharmacy.PatientEducCouns2009;75:283–9.
[43]MardbyAC,AkerlindI,HedenrudT.Generalbeliefsaboutmedicinesamong doctorsandnursesinout-patientcare:across-sectionalstudy.BMCFamPrac 2009;10:35.
[44]MoriskyDE,GreenLW,LevineDM.Concurrentandpredictivevalidityofa self-reportedmeasureofmedicationadherence.MedCare1986;24:67–74. [45]SungJC,NicholMB,VenturiniF, BaileyKL,McCombsJS,CodyM.Factors
affectingpatientcompliancewithantihyperlipidemicmedicationsinanHMO population.AmJManagCare1998;4:1421–30.
[46]GeorgeCF, PevelerRC,HeiligerS,ThompsonC.Compliancewithtricyclic antidepressants:thevalueoffourdifferentmethodsofassessment.BrJClin Pharmacol2000;50:166–71.
[47]NatarajanN,PutnamRW,YipAM,FrailD.Familypracticepatients’adherence tostatinmedications.CanFamPhysician2007;53:2144–5.
[48]BerrySD,QuachL,Procter-GrayE,KielDP,LiW,SamelsonEJ,etal.Poor adherencetomedicationsmaybeassociatedwithfalls.JGerontolSeriesABiol SciMedSci2010;65:553–8.
[49]KockN.WarpPLS2.0usermanual.Laredo,Texas:ScriptWarpSystems;2011. [50]TomarkenAJ,WallerNG.Structuralequationmodeling:strengths,limitations,
andmisconceptions.AnnuRevClinPsychol2005;1:31–65.
[51]GefenD,StraubD,BoudreauM-C.Structuralequationmodelingand regres-sion:guidelinesforresearchpractice.CommAssocInformSyst2000;4. [52]BaronRM,KennyDA.Themoderator–mediatorvariabledistinctioninsocial
psychologicalresearch:conceptual,strategic,andstatisticalconsiderations.J PersSocPsychol1986;51:1173–82.
[53]SkrondalA,Rabe-HeskethS.Generalizedlinearlatentandmixedmodelswith compositelinksandexplodedlikelihoods. In:BiggeriA,DreassiE,LagazioC, MarchiM,editors.Statisticalmodelling.Firenze:FirenzeUniversityPress; 2004.p.27–39.
[54]ColtmanT,Devinney T,MidgleyD,Venaik S.Formative versus reflective measurement models:twoapplicationsofformativemeasurement.J Bus Res2008;61:1250–62.
[55]WoldH.NonlinearIterativePartialLeastSquares(NIPALS)modeling:some currentdevelopments. In:KrishnaiahP,editor.MultivariateAnalysis.New York:AcademicPress;1973.p.383–407.
[56]WoldH.Pathmodelswithlatentvariables:theNIPALSapproach. In:Blalock HM,AganbegianA,BorodkinFM,BoudonR,CapecchV,editors.Quantitative Sociology:InternationalPerspectivesonMathematicalandStatisticalModel Building.NewYork:AcademicPress;1975.p.383–407.
[57]WoldS,RuheA,WH.Collinearityprobleminlinearregression.Thepartial leastsquares(PLS)approachtogeneralizedinverses.SIAMJSciComputStat 1984;5:735–43.
[58]GebauerJ,KlineDM,HeL.Passwordsecurityriskversuseffort:anexploratory studyonuser-perceivedriskandtheintentiontouseonlineapplications.J InformSystApplRes2011;4:52–62.
[59]MaikranzJM,SteeleRG,DreyerML,StratmanAC,BovairdJA.Therelationship of hope and illness-related uncertainty to emotional adjustment and
adherenceamong pediatricrenalandlivertransplantrecipients.JPediat Psychol2007;32:571–81.
[60]ByrneMK,DeaneFP.Enhancingpatientadherence:outcomesof medica-tion alliance training on therapeutic alliance, insight, adherence, and psychopathology withmental health patients. Int J MentHealth Nurs 2011;20:284–95.
[61]SobelME.Asymptoticconfidenceintervalsforindirecteffectsinstructural equationmodels.SociolMethodol1982;13:290–312.
[62]BarolettiS,Dell’OrfanoH.Medicationadherenceincardiovasculardisease. Circulation2010;121:1455–8.
[63]ByrneM,WalshJ, MurphyAW.Secondary preventionof coronaryheart disease: patient beliefs and health-related behaviour. J Psychosom Res 2005;58:403–15.
[64]Osterberg L, Blaschke T. Adherence to medication. N Engl J Med 2005;353:487–97.
[65]AmmassariA,MurriR,PezzottiP,TrottaMP,RavasioL,DeLongisP,etal. Self-reportedsymptomsandmedicationsideeffectsinfluenceadherencetohighly activeantiretroviraltherapyinpersonswithHIVinfection.JAcquirImmune DeficSyndr2001;28:445–9.
[66]Marinker M. Personal paper: writing prescriptions is easy. Brit Med J 1997;314:747–8.
[67]EkmanI,AnderssonG,BomanK,CharlesworthA,ClelandJG,Poole-WilsonP, etal.Adherenceandperceptionofmedicationinpatientswithchronicheart failureduringa five-yearrandomisedtrial. PatientEduc Couns2006;61: 348–53.
[68]RobertsonD,KellerC.Relationshipsamonghealthbeliefs,self-efficacy,and exercise adherencein patientswith coronaryarterydisease.HeartLung 1992;21:56–63.
[69]SchweitzerRD,Head K, Dwyer JW.Psychological factors and treatment adherencebehaviorinpatientswithchronicheartfailure.JCardiovascNurs 2007;22:76–83.
[70]JoekesK,VanElderenT,SchreursK.Self-efficacyandoverprotectionare relatedtoqualityoflife,psychologicalwell-beingandself-managementin cardiacpatients.JHealthPsychol2007;12:4–16.
[71]GilboyMB.Multiplefactorsaffectdietitians’counselingpracticesforhigh bloodcholesterol.JAmerDietAssoc1994;94:1278–83.
[72]SullivanMD,LaCroixAZ,RussoJ,KatonWJ.Self-efficacyandself-reported functionalstatusincoronaryheartdisease:asix-monthprospectivestudy. PsychosomMed1998;60:473–8.
[73]OhlssonH,LynchK,MerloJ.Isthephysician’sadherencetoprescription guidelinesassociatedwiththepatient’ssocio-economicposition?Ananalysis ofstatin prescriptioninSouthSweden. J EpidemiolCommunity Health 2010;64:678–83.
[74]StormA,AndersenSE,BenfeldtE,SerupJ.Onein3prescriptionsarenever redeemed:primarynonadherenceinanoutpatientclinic.JAmAcadDermat 2008;59:27–33.