Impact of pictograms on medication adherence : A systematic literature review

(1)

Review Article

Impact of pictograms on medication adherence: A systematic literature review

Hege Sletvold

^a,

*, Lise Annie Bjørnli Sagmo

^a

, Eirik A. Torheim

^b

aFacultyofNursingandHealthSciences,NordUniversity,Norway

bDepictAS,Oslo,Norway

ARTICLE INFO Articlehistory:

Received1July2019

Receivedinrevisedform17December2019 Accepted30December2019

Keywords:

Adherence Compliance Pictograms Medicines

Medicationcounselling

ABSTRACT

Objective:Theaimofthissystematicreviewwastoinvestigatethepotentialeffectofpictogramson patientadherencetomedicationtherapies.

Method: PubMed, MEDLINE, Embase, CINAHL, and CENTRAL were searched for relevant articles.

Experimentalstudiestestingtheuseofpictogramsinpatientcounsellingregardingmedicationtherapy, whichquantitativelymeasuredadherence,wereincluded.

Results:Seventeenstudieswereidentifiedthatfulfilledourinclusioncriteria.Thesewereheterogeneous withrespecttostudysetting,populationsize,andthemedicationregimentested.Allthestudieshad methodologicalqualitylimitations.Thepictograminterventionsdifferedwithrespecttocomplexity, interventionlength,andthemeasuredadherenceoutcome.Tenstudies(58.8%)reportedastatistically significant effect, of the pictogram intervention in question, on patient adherence to medication therapies.Ofthese,80%involvedpopulationsatelevatedriskfornon-adherence.

Conclusion and practice implications: Pictograms used in combination with written and/or oral informationcanhaveapositiveimpactonpatientpopulationsthatarehighlyatriskfornon-adherence whencounselledontheproperuseofmedicines.

Contents

1. Introduction ... 1095

2. Methods ... 1096

3. Results ... 1099

4. Discussionandconclusion ... 1100

4.1. Discussion ... 1100

4.2. Implications ... 1102

4.3. Limitations ... 1102

Funding ... 1102

DeclarationofCompetingInterest ... 1102

CRediTauthorshipcontributionstatement ... 1102

References... 1102

1.Introduction

Patient adherence to medication therapies is a primary determinantoftreatmentsuccess,andmedicationnon-adherence

is a common and recognized problem in health care [1–3].

Adherence is defined as “the extent to which a person’s behaviour—takingmedication,followingadiet,and/orexecuting lifestyle changes—corresponds with agreed-upon recommenda- tionsfromahealthcareprovider”[1].However,manypatientsdo not follow treatmentrecommendationsfor different(andcom- plex) reasons[1,2]. The WorldHealth Organization (WHO) has definedfive“dimensions”thataffectadherence:Thepatient,the therapy,thehealthsystem,thecondition,andthesocioeconomic

* Corresponding author at: Faculty of Nursing and Health Sciences, Nord University,PB324,N-7501,Stjordal,Norway.

E-mailaddress:Hege.sletvold@nord.no(H.Sletvold).

https://doi.org/10.1016/j.pec.2019.12.018

ContentslistsavailableatScienceDirect

Patient Education and Counseling

j o u r n a l h o m e p a g e : w w w . e l s ev i er . c o m / l o c a t e/ p a t e d u c o u

(2)

environment[1].Medicationnon-adherencemayresultinadverse healthoutcomesandincreasedhealthcarespending[1,3].

Medical informationsupportsmedicationadherencethrough facilitatingthecorrectuseofmedicinesandimprovingthelevelof understandingwhilesimultaneouslygeneratingpositiveattitudes towards treatment [1]. Traditionally, information for patients regarding correct medicine use (in Europe) involves verbal counsellingbyhealthcarepersonnel(HCP),whichissupplemented bywritteninformationintheformofapatientinformationleaﬂet (PIL)providedbythemedicinemanufacturer[4].However,several shortcomingsapplytothis approach:ThePILshavenoconsent standards,areknownforsmallprintandlengthytextswrittenat anadvancedreading level,andcontaingeneralisedratherthan personalisedinformation[5–7].

Notably,picturesorpictogramsthatgraphicallyillustrateand/

or emphasise instructions for the correct use and storage of medicines can facilitate communication between HCPs and patients.Apictogramcanbedeﬁnedasapictureorsymbolthat represents a word or phrase.A literature reviewreported that including pictograms in patient counselling could reduce the frequencyofmedicationdosingerrorsrelatedtotheadministra- tionofliquidmedications[8].Moreover,pictogramsincombina- tionwithspokenand/orwrittenmedicinalinstructionshavebeen demonstrated to enhance the visual attention,comprehension, andrecallofmedicationinstructionsprovided[8–10].Ithasalso beenreportedthatpatientsprefertohavepictogramsandwritten ororalinformationusedincombination.Thisisespeciallytruefor patientswithlowhealthliteracy,aswellaselderlypatients[10].

Patient-centricdevelopment,aswellastestingandvalidationin distinct patient populations, is of importance to the utility of pictogramsusedinmedicationcounselling[11–13],herebycalled pharmaceuticalpictograms.

Previousreviewshaveemphasisedtheroleofpharmaceutical pictogramsonpatient comprehension and recall of medication instructions,buttheseincludeonlyasmallnumberofstudiesthat report an effect onpatient adherence [8–10]. The aim of this

systematic review was to investigate the potential effect of pictograms on patient adherence to medication therapies.

Additional objectives of the studywere todetectand describe pictogram interventionsthat positivelyaffectadherence and to evaluatethepotentialcontributionofpictogramstotheefﬁcacyof complexinterventions.

2.Methods

Thisisasystematicliteraturereviewperformedaccordingto thePreferredReportingItemsforSystematicReviewsandMeta- Analyses(PRISMA)Statement[14],wherethedatabasesPubMed, MEDLINE,Embase,CINAHL,andtheCochraneCentralRegisterof ControlledTrials(CENTRAL)weresearchedforrelevantstudies.

Aninitial search in PubMedwasconducted in December2017 withthefollowingsearchstrategy:(pictogramORpictographOR pictureORpictorialORgraphicsORvisualsORiconORsymbol) AND (adherence OR compliance OR concordance), limited to publicationdatesfrom1997/01/01to2017/12/31,andEnglishas the publicationlanguage. PubMed,MEDLINE, Embase,CINAHL, andCENTRAL were searchedin May2018 using the following search strategy (referred to as the second search strategy):

(pictogram OR pictograph OR pictorial OR picture), combined with(AND)thefollowingMeSHTerms:(medicationadherence OR patient adherence ORpatient compliance). These searches were limited to English language studies published between 1997/01/01to2018/05/31.Additionalﬁlterswere“peerreviewed” or “academic” journalsfor the MEDLINEandCINAHL searches, respectively.

ThePopulation,Intervention,Comparison,andOutcome(PICO) ofthestudywere:patientsonmedication(Population),pictogram (s)hadtobeinvolvedinanexperimentalstudydesign(Interven- tion),nospeciﬁccriteriaforthecomparison(C),andthestudies had to quantitatively measure adherence (Outcome). To be included,thestudiesalsohadtoreportonoriginaldataandbe publishedinapeer-reviewedjournal.

Fig.1.Studyﬂowdiagram.

(3)

Table1

Characteristicsandsummaryofincludedstudies.

Study reference

Study setting, country

Study design^a

N(loss to follow- up)

Studypopulation^b Intervention(s)^c Control Adherenceeffect measurement^d

Result

Braichetal.

2011[33]

Clinic,India RCT, three arms

225 (87)

Lowliteracy patientson postoperative cataract medications(eye drops)

Educationonmedication use:

I1)Oralinformation combinedwith pictograms

I2)AsforI1,butpatients tookpictogramshome

Oral educationon medication use

Eyedropbottleamount measurementat baselineand28days aftersurgery

Bothintervention groupshadsigniﬁcantly improvedadherenceat day28(p<0.001),as judgedbythe percentageofeyedrops used

Chanand Hassali 2014[34]

Outpatient pharmacy, Malaysia

RCT, three arms

126 (16)

Patientson antihypertensive andantidiabetic medications

I1)Medicationlabels usingenlargedfonts I2)Medicationlabels incorporatingpictograms

Regular-size text medication labelswithout pictograms

MMAS-8atbaselineand 4weeksafter intervention

Nosigniﬁcanteffectson adherencebetween studygroups

Dowseetal.

2014[35]

Clinic,South Africa

RCT 116

(52)

HIVpatientsnew toARVT

PILcontainingtextand pictograms

Standardcare HIVTreatment AdherenceSelf-efﬁcacy Scale(HIV-ASES,0–10 scale)atbaselineand1, 3,and6monthsafter intervention

Nosigniﬁcanteffectson adherenceself-efﬁcacy betweenstudygroups

Dowseand Ehlers 2005[24]

Outpatient clinic,South Africa

RCT 87(0) Patientsonshort- course

antibacterial medications

Medicationlabels incorporatingpictograms

Text-only medication labels

Pillcount/volumetric measurementandself- reportedadherence3–5 daysafterintervention, reportedasacombined adherenceresultin%

Signiﬁcantly(p<0.01) higheradherenceinthe interventiongroup(89.6

%)comparedtothe controlgroup(71.5%) Holzheimer

etal.1998 [36]

Outpatient, Australia

RCT, fourarms

80(24) Children(2–5 yearsofage)on prophylacticanti- asthmatic medications

Interventionsusedin asthmaeducationof children;

I1)Asthmavideotapeand asthmabookincluding pictograms

I2)Asthmavideotapeand unrelatedbook I3)Unrelatedvideotape andasthmabook includingpictograms

Unrelated videotapeand unrelated book

Parentdiarystartingat1 monthpre-intervention andcontinuinguntil3 monthspost- intervention

Nosigniﬁcanteffects betweenintervention groupsandcontrol regardingnon- compliancedays

Kalichman etal.2013 [29]

Outpatient, USA

RCT, three arms

446(45 forpill count)

Lowhealth literacyHIV patientsonARVT

I1)Adherencecounselling includingwritten informationwith pictograms,adherence toolofchoice I2)Standardadherence counsellingincluding writteninformationwith illustrationsandcomic strips,adherencetoolas pillbox

General health improvement counselling

HIVRNAviralloadat baselineand9months afterintervention,and monthlypillcounts frominterventionstart for9months

Signiﬁcantlygreater undetectableHIVviral loadsforpatientswith marginalliteracyinboth interventiongroups comparedtocontrol.

Patientswithlower healthliteracy demonstratedno signiﬁcanteffectsof interventions.

Kripalani etal.2012 [37]

Primarycare clinic,USA

RCT, fourarms

420 (20)

Patientswith coronaryheart disease

I1)Reﬁllreminder postcards

I2)Medicationschedules includingpictograms I3)CombinationofI1and I2

Usualcare Electronicpharmacy reﬁllrecordsreportedas CMGfor1yearoffollow- upafterinterventions

Mansoor and Dowse 2006[23]

Primarycare clinic,South Africa

RCT, three arms

127(7) Lowhealth literacypatients onARVT

I1)PILincluding pictograms I2)PILwithout pictograms

Usualcare(no PIL)

Pillcountandself- reportedthrough questionnaires approximately14days afterintervention

Signiﬁcantly(p<0.05) improvedadherence bothbasedonpillcount andquestionnaire,inthe I1groupcomparedtoI2 andcontrol

Mohanetal.

2014[38]

Safetynet clinic,USA

RCT 208(8) Patientson antidiabetic medications

Personalisedmedication scheduleincluding pictograms(PictureRX^TM)

Usualcare with handwritten listof medication

ARMS(8items) approximatelyoneweek afterenrolment

Monroe etal.2018 [39]

HIVclinic, USA

RCT 46(4) AdultHIVpatients onARVTand antihypertensive and/or antidiabetic medications

Personalisedmedication scheduleincluding pictograms(PictureRX^TM)

Usualcare withregular discharge medication list

Electronicpharmacy reﬁllrecordsreportedas MPRcalculatedfrom180 dayspre-intervention and180dayspost- intervention

Murrayetal.

2007[27]

Ambulatory care practice, USA

RCT 314

(44)

Heartfailure patients50years ofageusing

Pharmacistmulti-level interventionincluding writtenmedication informationand

Usualcare MEMS,MPRbyusing prescriptionrecordsand self-reportedthrough questionnaires.

Signiﬁcanteffecton overalladherence(%of prescribedmedication taken)between

(4)

All search results were exported to EndNote X8.1 software.

ResultsfromtheinitialPubMedsearchwereevaluatedseparately bythethree reviewauthorsaccordingtotheinclusion criteria.

First,theresults werescreenedbyreadingthearticletitlesand

excludingarticlesthatwerenotrelevantaccordingtotheinclusion criteria.Next,thestudyabstractswereevaluated,andnon-relevant articleswereexcluded.Finally,thefull-textarticlesselectedbyall three authors were collected and assessed for their relevance Table1(Continued)

Study reference

Study setting, country

Study design^a

N(loss to follow- up)

Studypopulation^b Intervention(s)^c Control Adherenceeffect measurement^d

Result

cardiovascular medications

medicationlabelsthat containedpictograms,9- monthactive

interventionperiod

Measurementsduring interventionperiod(9 months)and3months post-intervention

interventionandcontrol group(10.9%difference, 95%CI5.0–16.7%).The effectdissipatedinthe post-intervention period.Nosigniﬁcant effectsonadherence betweenstudygroups whenmeasuredbyself- reporting

Negarandeh etal.2013 [30]

Diabetic clinic, secondary carelevel, Kurdistan

RCT, three arms

135(8) Lowhealth literacypatients withtype2 diabetes

Educationalmedication interventionsconsisting of:

I1)Theteach-back method I2)Pictograms

Usualcare MMAS-8atbaselineand 6weekspost- intervention

Signiﬁcantly(p<0.001) higheradherenceinthe interventiongroups(I1 6.73,I27.03)compared tothecontrolgroup (3.63),butnot signiﬁcantbetween interventiongroups Ngohand

Shepherd 1997[25]

Outpatient, Cameroon

RCT, three arms

78(0) Illiteratepatients onantibacterial medications

Educationalmedication interventionsconsisting oforaleducationin additionto:

I1)Pictogramsandan

“advancedorganiser”

I2)Pictograms

Usualcare Pillcountonorafterthe fourthdayof

intervention,presented in%(pillcount adherenceratio)

Signiﬁcantly(p<0.05) higheradherenceinthe interventiongroups (I1=94.6%andI2=89.6

%)comparedtothe controlgroup(77.5%) Okonkwo

etal.2001 [26]

Outpatient, Nigeria

RCT,three arms

632 (180)

Children(0.5–5 yearsofage)with malariaon chloroquinesyrup

Medicinedispensing informationgivenas:

I1)PILincluding pictograms I2)PILincluding pictogramsandverbal instructions

Medicine dispensing without information

Volumetric

measurementandself- reportedbya questionnaire48hours afterintervention.

Resultsgivenas combinedresultsand deﬁnedadherenceas non-compliant,partial compliant,orfully compliant

Signiﬁcantly(p<0.001) higherproportionof fullycompliantchildren intheintervention groups(I151.9%,I273.3

%)comparedtothe controlgroup(36.5%)

Shetetal.

2014[32]

Clinic,India RCT 631 (98)

AdultHIVpatients initiatingARVT

Adherencesupportbya mobilephone

interventiononceweekly forthestudyduration.

Theintervention included:

a)interactivevoice responsecalls b)neutralpicture (pictogram)messaging services

Usualcare HIVRNAviralload(time tovirologicalfailure) andpillcounts measuredatbaseline andatweeks4,8,and 12,andthenevery12 weeksuntilweek96

Yinetal.

2008[28]

Hospital, USA

RCT 245

(18)

Parentsor caregiversof children(30days to8yearsofage) onliquid medications(daily doseand14days oftherapy)

Medicationcounselling includingmedication instructionsheetswith plainlanguage, pictograms(HELPix),and teach-back

Usualcare Self-reportedadherence byinterviewatbaseline and3–5daysafter medicinedispensing, reportedasnon- adherencein%(>20% deviationofthe prescribeddose)

Signiﬁcantly(p<0.05) higheradherenceinthe interventiongroup(9.3

%non-adherent) comparedtothecontrol group(38%)

Zerafaetal.

2011[18]

Hospital, Malta

RCT (reported by authorsas case- control)

86(6) Cardiacsurgery patientson medications

Pharmacistintervention consistingofmedication counsellingwithwritten medicationinformation sheetsincluding pictograms

Usualcare Patientcompliance questionnaire8weeks aftersurgerydischarge

Signiﬁcantly(p<0.001) highercomplianceinthe interventiongroup(88.2

%)comparedtothe controlgroup(66.4%)

aRCT,randomisedcontrolledtrial.

b ARVT,antiretroviraltherapy;HIV,humanimmunodeﬁciencyvirus.

cI1,interventioninstudyarmone;I2,interventioninstudyarmtwo;I3,interventioninstudyarmthree;PIL,patientinformationleaﬂet.

d MMAS-8,8-itemMoriskyMedicationAdherenceScale;CMG,cumulativemedicationgap;ARMS,AdherencetoReﬁllsandMedicationsScale;MPR,medicationpossession ratio;MEMS,MedicationEventMonitoringSystem.

(5)

relativetotheinclusioncriteria.Anydisagreementsregardingthe eligibilityofstudieswerereconciledattheﬁnalstepbydiscussion andconsensus.

ResultsfromthesecondsearchstrategyinPubMed,Embase, CINAHL,MEDLINE,andCENTRALwereinitiallyevaluatedbyoneof the review authors (H.S.) by removing article duplicates and evaluatingtitlesandabstractsasdescribedabove.Additionally,the referencelists of theidentified studies werehand searched to retrieveadditionalrelevantarticles.Finally,theretrievedfull-text articleswereindependentlyreviewedbytwoofthethreeauthors accordingtotheinclusioncriteria.Incasesofuncertaineligibility, allthreeauthorsreadthearticle,andconsensusonwhetherornot to include the article was reached by discussion. In total, the reviewersdiscussedinclusionforninestudies.Studydesignand theoutcomemeasurement(adherencemeasure)werereasonsfor discussionfor five and four studies, respectively. A study flow diagramisprovidedinFig.1.

Theincludedstudieswereindependentlyanalysedbytwoof theauthors,anddatawereextractedtopresentthekeyfeaturesof theinterventionstudiesandreﬂectonpointsof differencethat couldaffecttheinterpretationofthepictogramintervention.The following variables wereextracted: study setting, study design (deﬁnedasrandomisedcontrolledtrial[RCT]ornon-randomised study[NRS]),number ofstudyparticipants,shortdescriptionof studyparticipants,typeofintervention(s)withemphasisonthe pictogram contribution, adherence effect measure(s), time of adherencemeasurement,andadherenceoutcome.

TheRCTstudieswereassessedbytwooftheauthorsforriskof biasusingtheCochraneCollaboration’stoolforassessingriskof biasinrandomisedtrials[15].RCTswithahighriskofbiasinfour ormoredimensionswereexcluded.TheNRSswereassessedby twooftheauthors,whodescribedpre-andpost-interventionrisk ofbiasaccordingtoanabridgedversionofROBINS-I—a toolfor assessing the risk of bias in non-randomised studies of interventions[16].NRSswithacriticalorseriousriskofbiasbothatpre- andpost-interventionwereexcluded.

The heterogeneity of the studies with respect to patients, pictograminterventions,andadherenceoutcomemeasurements precludedameta-analysis.Consequently,theresultsofthisreview arepresentednarratively.

3.Results

Ourdatabasesearchesidentiﬁed1,283studies,ofwhich 358were duplicateresults.Atotalof896studieswereexcludedbasedontheir titleorabstract,whileafurthertwelvestudieswereexcludedupon full-textreview.Seventeenstudieswereincludedforanalysisinthis literaturereview,tenofwhichwerepublishedinthepast10years.

ThestudieswereinitiallycategorisedaseitherRCT(n=18)orNRS (n=5),withpictogramsbeingpartofaninterventiontoincrease medicationadherenceorcompliance.TheNRSstestedpictograms inapre-postinterventionstudydesign,butwereallexcludedfrom theﬁnalanalysisduetopoorstudyquality.Oneotherstudy,which was reportedlya case-controlstudy, was included [18] since it tested an intervention with the aim of increasing medication adherence, andwasthereforejudged by theauthorsofthisreviewto beanRCTwithanexperimentalstudydesign.

The included studies were heterogeneous regarding study setting,thenumberofparticipants,studypopulation,thechoiceof interventionsandcontrols,andtheadherenceeffectmeasurement used(seeTable1).Regardinggeographiclocation,sixstudieswere conductedintheUSA,while twowereconductedin India,and threeintheRepublicofSouthAfrica. Theremainingsix studies were performed in Cameroon, Malaysia, Australia, Kurdistan, Nigeria,andMalta,respectively.Studieswereconductedbothin primary and secondary healthcare settings. The number of

participantsineachstudy(n=numberofparticipantsatthestart oftheintervention)variedbetween46[39]and632[26],withthe median numberof participantsbeing135[30]. Atotal of3,995 patients were includedacross all studies combined. The study population was diverse in terms of age, clinical disorders, treatment regimens, and the level of health literacy. The most frequently targeted medication therapies were antiretroviral therapy (ﬁve studies) and cardiovascular medications alone or in combination withantidiabetic medicines (four studies). The remainingstudiestargetedantidiabeticmedicines(twostudies), antibacterial medicines (two studies), anti-asthmaticmedicines (onestudy),chloroquinesyrup (onestudy), cataractmedication (one study), and general liquidmedications (onestudy). While pictogramswereinvolvedininterventionsregardingmedication therapyinallincludedstudies,thereweresubstantivedifferences between the studies regarding the intervention complexity, healthcare personnelinvolvedin theinterventions, intervention length,andtheadherenceoutcomesmeasured.Asaresult,there wasinsufﬁcientcommongroundforquantifyingtotaldifferences between intervention and control groups or estimating pooled effectsizesforanalysisacrossstudiesregardingtheoveralleffect onadherence.

Oftheincludedstudies,10studies(58.8%)reportedastatistically significanteffectofpictograminterventionsonpatientadherenceto medicationtherapies.Allstudieswereperformedinhospital,clinic, or outpatient settings. The studies differed with respect to the medicationtherapiesselected,numberofmedications,aswellas whetherthetreatmentwasshort-termorforchronicuse.Insome cases,sufficientdetailsregardingthemedicationtherapieswerenot provided.Forexample,thestudybyMansoorandDowseinvolved medicinesusedinantiretroviraltherapy(ARVT)butdidnotspecify whichmedicineswereadministered[23].Incontrast,themedica- tion therapies used in studies by Dowse and Ehlers, Ngoh and Shepherd, and Okonkwo et al. were well defined (amoxicillin, phenoxymethylpenicillin, and co-trimoxazole; ampicillin, sulfa- methoxazole-trimethoprim,metronidazole,andtetracycline;chlo- roquine syrup, respectively) [24–26]. Some studies described polypharmacyin theirstudypopulation(e.g.,Murrayet al.,and Zerafa et al. [18,27]), whereas others did not provide any informationregardingothermedicinesused[28].

Insixofthetenstudiesdemonstratingasigniﬁcanteffectofa pictogram-basedintervention,theinterventionswerecomplexand involved both pictogramsand medicationcounselling combined withadherencetoolsorteach-back[18,25,27–30].Inotherstudies, plaininterventionsusingpictogramsinpatientmedicationinfor- mationandinstructionleaﬂets[23,26],oronlabels[24],alsoproved tobeeffectiveinincreasingpatientadherence.

Variousmethodswereusedtomeasureadherence,including pillcount,volumetric measurements,electronicpharmacyrefill records, viralload,self-reportedadherencescales,self-reported adherencebyvignettesorinterviews,andtheuseofmedication eventmonitoringsystems(MEMS).Consequently,theadherence definitionsandoutcomeeffectswerenotreportedinaconsistent manner.Forinstance,inastudybyDowseandEhlers,adherence was determined by pill count or volumetric measurement of antibacterial tablets or suspensions, respectively. Additionally, patientsinthisstudyreportedonadherenceusingaquestionnaire, withatotaladherencescorebeingcalculatedandconvertedintoa percentage [24]. In contrast, Ngoh and Shepherd measured adherenceusingpillcountsonly,andresultsweregivenasapill count adherence ratio [25]. Subjective self-reported adherence effectmeasureswereusedin11 ofthestudiesincludedinthis review[18,23,24,26–28,30,34–36,38].Anexampleisthestudyby Negarandehetal.,whichutilisedthe“8-itemMoriskyMedication AdherenceScale” (MARS-8)[30].Thisstudyreportedsignificant effectonadherence usingmedicationeducationalinterventions

(6)

consisting of pictograms [30]. Such a questionnaire defines adherenceaccording to thespecific scale used, rather than an adherence percentage. Zerafa et al. used a novel self-made questionnaire named “Assessing Patient Compliance” that reportedeachpatient’smeanpercentagecomplianceascalculated bythetotal numberof correct andincorrectanswers [18]. The study found a significant effect on patient compliance upon comparinga pharmacist interventionof medicationcounselling andwrittenmedicationinformationsheets(includingpictograms) tostandard care [18]. Mixed methods were used in six of the studies [23,24,26,27,29,32], of which five studies reported a significantpictogramintervention effect[23,24,26,27,29]. When usingmixedmethods,adherenceoutcomeswereeitherreported separately([23,29]orincombination[24,26,27].

Among thetenstudies reportingsigniﬁcanteffects ofpicto- gram-based interventions, heterogeneity was observed in the choiceofpatientgroup.ExamplesofpatientgroupsincludeHIV patientsonantiretroviraltherapy(ARVT),patientsonantibacterial medications, cardiac and diabetic patients, and children on medication(see Table 1). Patientgroups known to experience challengeswithadherence,suchaspatientpopulationswithlow (health) literacy [23–25,29,30,33] and children and/or their caregivers[26,28],wereinvolvedineightofthetenstudies(80%).

All included RCTs were assessed according to the Cochrane Collaboration’stoolforassessingriskofbiasinrandomisedtrials [15].Thisinvolveda thoroughanalysisoftheriskofbiasacross seven dimensions: 1) Random sequence generation (selection bias);2)Allocationconcealment (selectionbias); 3)Blindingof participants and personnel (performance bias); 4) Blinding of outcomeassessment(detectionbias);5)Incompleteoutcomedata (attrition bias); 6) Selective reporting (reporting bias); and 7) Otherbias.Theindividualresultsofthisanalysisareprovidedin Fig. 2. Fig. 3 presents the cumulated results as stacked bars, presentingtherelativefrequenciesof studiesdeemedtohavea low,unknown,orhighriskofbiasineachofthesevendimensions.

Ourresultsdemonstratethat,withoutexception,alloftheRCTs identifiedandincludedinthisreviewhadahighriskofbiasinat least one dimension. Notably, they all had a high risk of performancebias,reflecting insufficientblindingof participants and/orpersonnel.In threeofthestudies[27,29,36], researchers were blinded to compensate for difficulties in the blinding of participants and care providers to the use of pictogram-based interventions.However,whileweconsiderthisanimprovement overnoblindingatall,thefactthatparticipantallocationisknown totheparticipantsthemselves,aswellasthecareproviders,must beconsideredtoresultinahighriskofperformancebias.

Thesecondgreatestcontributortostudybiasinthisreviewwas theinsufﬁcientblindingofoutcomeassessments,leadingtoahigh riskofdetectionbiasin10of17studies.Considerableriskofbias was also identiﬁed in the dimensions of selective reporting

(4 studies with high risk of bias) and allocation concealment (3studieswithhighriskofbias).Ahighriskofattritionbiaswas onlyobservedinonestudy[22].Themajorityofstudieshadalow risk of bias regarding random sequence generation, while six studieshadanuncertainlevelofriskinthisdimension.

Fivestudieswereexcludedfromthereviewfollowingafull-text reading, since they did not quantitatively measure adherence [40–44], while one study was only a study protocol [45].

Additionally, six studies were excluded because of poor study quality.OneRCTstudywasexcludedduetohighriskofbiasinfour ormoredimensions[22].Fiveoftheexcludedstudieswerenon- randomised studies [17,19–21,31]. A critical risk of pre- intervention biaswas observed in allﬁve studies.In the study byGazmararianetal.,theinterventionandcontrolgroupswere locatedatdifferentstudysites[31].Moreover,studiesbyHawkins et al. and Martin et al. used convenience sampling of study participants[20,21],while thestudybyVetteretal.usedtime- sequentialsampling[17].Notably, thestudybyRodrigueset al.

lackedinformationtojudgetheriskofbiasinstudyparticipant selection[19].Allﬁvestudieswereobservedtohaveacriticalor serious risk of post-intervention bias as well as bias in the measurementofoutcomes[17,19–21,31],whileGazmararianetal.

andRodriguesetal.alsohadbiasintheselectionofreportedresults [19,31].Twoofthestudieswerepilotstudieswithasmallnumber ofparticipants[20,21].

4.Discussionandconclusion 4.1.Discussion

Inthisreview,wehaveassessedstudiesdesignedtoevaluate theeffectofpictogram-basedinterventionsonpatientadherence Fig.2. RiskofbiasanalysisoftheincludedRCTs.+,lowriskofbias;?,unclearriskofbias;-,highriskofbias.

Fig.3.CumulativeriskofbiasobservedintheincludedRCTs.

(7)

tomedications.Fromasizeableanddiverseevidencebase,various uses of pictograms and different measures of adherence were investigated. Heterogeneity in the design and conduct of the included studies precluded any meta-analysis of observed pictogrameffects.However,ofthe17studiesincluded,10studies (58.8 %) reported a statisticallysigniﬁcanteffect of pictogram- containing interventions on patient adherence to medication therapies.

The pictograms utilised in the studies varied, though the majorityofstudyinterventionsutilisedpictogramsincombination withverbalmedicationcounsellingortext-basedinstructionsof medication therapy. A review by Katz et al. concluded that pictogramsusedtocomplementtextualororalinformationwere moreeffectiveatimprovingpatients’understandingofcorrectuse ofmedicinesthanpictogramsalone[10].Thecurrentreviewpoints toapossibleeffectofpictogramsincombinationwithoralortext- basedmedicationinformation or counsellingwhen it comesto improvingpatientadherence.

The absenceof significant effects onpatient adherence was common among many of the reviewed studies. We primarily attributethis toinsufficient samplesizes. Measuringadherence has proven difficult in many cases, and there is a need for standardisationofreproducibleadherencemeasures.Amongthe studies included in this review, self-reported adherence by questionnaireorinterviewwas themostconsistentlysuccessful measureforidentifyingdifferencesinadherence.However,other methodologicalaspectsmayhavealsocontributedtothelackof significantresultsinmanyofthestudies.Notably,wehavefound sufficientevidenceinourreviewtosustaintheassumptionthat pictogram-based interventions may indeed serve to improve patients’adherencetomedicationtherapies(seeTable1).

Thecomplexityofinterventionslimitsourabilitytointerpret theexactcontributionofthepictogramsonmedicationadherence.

Forinstance,thestudybyKalichmanetal.utilisedpictogramsin combinationwithadherencecounsellingandadherencetoolsto enhance antiretroviral therapy [29]. While the study reported statisticallysignificanteffectonadherence,itcouldnotdetermine theexactcontributionofpictogramsontheobservedoutcome.In theRCTstudybyNegarandehetal.,aninterventionconsistingof diabetesmedicationeducationbyanursecombinedwithteach- backor pictograms (in two separate intervention groups), was tested against standard care [30]. Significant differences in medicationadherencewereobservedbetweentheintervention groupsand thecontrolgroup,thoughnosignificantdifferences were observed between the two intervention groups [30]. In contrast,theRCTby Mansooret al.testedPILswithor without pictograms in two different intervention groups compared to standard care. In this case, adherence to ARVT improved significantlywhenpatientsreceivedPILswithpictogramswhen comparedtotheothertwogroups[23].Totestthetrueeffectof pictograms, study designs must allow for comparison of two interventionswhere theuseofpictograms constitutestheonly difference.

Patient-relatedfactors may contribute tomedication adherence, since age, literacy, and cognitive function have all been identiﬁedasnegativelyimpactingadherence[1,2].Ineightofthe tenstudieswithsigniﬁcantinterventioneffects,patientpopula- tionsconstitutedindividualswithlowhealthliteracyorlowage.

DowseandEhlerstestedmedicinelabelsincorporatingpictograms on Xhosa African patients using antibacterial medications and observed aneffect of literacyonadherence when resultswere pooledforinterventionandcontrolgroups[24].Furthermore,the study by Kalichman et al. described an adherence counselling interventionincludingpictograms,inwhichaneffectwasfound among the marginal literacy HIV patients on ARTV, but with conﬂicting results among lower literacy patients [29]. In the

studies by Mansoor et al., Negarandeh et al., and Ngoh and Shepherd, theroleofliteracyinaffectingmedicationadherence was not tested, yet the study populations were selected with literacyasoneoftheselectioncriteria[23,25,30].Theseresults indicatethatpharmaceuticalpictogramsareparticularlybeneﬁcial topatientpopulationschallengedbylowlevelsofhealthliteracy.

Anotherfactorknowntoinfluenceadherenceisthenatureof thetherapy[1,10].Thestudiesinthisreviewwerehighlydiverse regardingtherapytypeandweredescribedwithvaryinglevelsof detail. Forinstance,in theRCTbyNegarandehetal.,adherence amongtype2diabeticpatientswasmeasured,thoughnodetails were provided regarding theprescribed therapies of the study participants [30]. Similarly, a number of the reviewed studies failed to describe the participants’ medication therapies in sufficient detail [18,23,28,29]. However, all of these studies described a significant effect of pictogram-based interventions on medicationadherence. Therapeutic aspects knownto affect adherenceincludefrequencyofdosing,complexityoftreatment, adverse effects of treatment, the patient's beliefs and attitudes regardingthetreatment,andmedicationeffectiveness[1,2].Itis notknowntowhatextentthesefactorshaveinfluencedtheresults ofthestudiesincludedinthisreview.

Severaldifferentmeasuresofadherencewereobservedinthe reviewedstudies.Agoldstandardadherencemeasureshouldbe userfriendly,highlyreliable,flexible,practical,cheap,andeasyto perform. Unfortunately, no single method exists, and a mixed methodapproachisthereforerecommended[46].Inthisreview, sixoutof17studiesusedamixedmethodapproachtomeasure adherence[23,24,26,27,29,32],andfiveofthesestudiesreporteda significant adherence effect [23,24,26,27,29]. Evidently, when using several methods of measurement for thesame outcome effect, theanalysisand interpretation of resultsbecomes more complex. Therefore, it is important to choose complementary rather than potentially conflicting measures; for example, a combinationofasubjectiveandanobjectivemeasurement,with suitabilityforthestudysetting,studyparticipants,andanyother practical considerations. Many studies in this review (n = 11) utilisedsubjectiveself-reportedadherenceeffectmeasures(e.g., questionnaires)toassesspictogrameffect.Werecommendusing consistentandvalidatedquestionnaires,intheinterestofenabling statistical comparisons across studies. Self-reported adherence scales can measure medication-taking behaviour, and can also identifyadherencebarriersand/orpatientbeliefsassociatedwith adherence[47].Whenmeasuringadherence inanyintervention study, selecting the most suitable questionnaire remains important.

The studies included in this review were generally of poor quality.Onechallengewithpictogram-basedinterventionsisthat studyparticipants,ingeneral,cannotbeblinded.Still,onlythree studiesdescribedresearchersbeingblindedtoparticipantalloca- tion[27,29,36],whileonlysixstudiesreportedusingmeasuresto blindtheoutcomeassessment[29,32,33,36–38].Furthermore,ﬁve studies did not describe the method used to create a random allocationsequence[23–25,33,36],andsevenstudieshadahigh risk ofselection biasdue toinsufﬁcient allocationconcealment [18,23–25,33–35]. Unfortunately, the observed challenges with studyqualitycomplicatedourassessmentofthereportedresults andprohibitedstatisticallysoundcross-studycomparisons.

While somepreviousreviews haveinvestigatedtheeffectof pictogramsonpatientadherencetomedications,theycovereda verylimitedevidencebase.ThereviewbyKatzetal.concludedthat pictogramsenhancepatients’understandingofmedication-taking behaviour,yetincludedonlytwostudiesthatassessedadherence [10].Furthermore,Chanetal.reviewedtheeffectsofpictograms whenassistingcaregiversinliquidmedicationadministrationand suggestedthat pictogramsmightreducedosingerrors, enhance

(8)

the comprehension and recall of medication instructions, and improveadherence[8].However,thisreview includedonlyﬁve studies, and only one study that measured adherence as an outcome[8].Barrosetal.aimedtoevaluatetheuseofpictograms ina healthcaresettingtoassesstheirpotentialeffectonpatient understanding and medical instruction compliance [9]. They concludedthatpictogramscanserveascommunicationtoolsto enhancevisualattention,comprehension,recall,andadherenceof providedinstructions,thoughtheydidnotanalysetheeffectson adherenceindetail[9].Additionally,thereviewbyNicolsonetal.

concludedthat thereissomeevidencethatwritteninformation can improve patients’ knowledge regarding medicines when comparedtonowritteninformation;however,due tothepoor qualityofincludedstudies,thereviewlackedrobustconclusions [48].Finally,areviewofreviewsbyRyanetal.,thatassessedthe effects of interventions that targeted healthcare consumers to ensuresafeandeffectiveuseofmedicines,foundlimitationsinthe methodologicalqualityofincludedstudies,bothatthereviewand studylevel[49].

4.2.Implications

Thepresentreviewdocumentsapossibleeffectofpictogramson patients’adherencetomedications,especiallywhencombinedwith writtenand/ororalmedicationinformationandutilisedonpatient populationsathighriskofnon-adherence.However,theheteroge- neityinstudydesignandquality,aswellasintheinterventionsand outcomesmeasured,preventedusfromconclusivelyassertingthat pictogramsareeffectiveinimprovingadherence.

We ﬁnd it critical that studies examining the effects of pictograms have a high-quality study design adhering to best practisesofinterventionstudies,anduseconsistentandvalidated outcomemeasuresofadherence.Assuch,thereremainsaneedfor morehigh-qualitystudies,aswellasasetofstandardisedtoolsand protocols—ideally open source—that are tailored to adherence studies.Moreover,studiesassessingthevalueofpharmaceutical pictogramswouldbeneﬁtfromestablishingbestpracticesinthe designanduseofthepictogramsthemselves.

4.3.Limitations

The present review includes studies from 1997 until 2018;

hence,studiesupto20yearsoldwereincluded[25,36].Studies performedbeforetheConsolidatedStandardsofReportingTrials (CONSORT)statementof2010[50],whichdefinesbestpracticein reportingRCTs,maybeofpoorerquality.Recentdevelopmentsin thehealthcaresector,andinthewayspatientsreceiveandaccess information, may also have rendered the older studies less relevant.Among the studies includedin this review, ten of 17 studies were conducted in 2010 or later. The level of patient adherencemajorlyimpactshealthandhealthcareexpenditureand, hence,adherenceservedasa naturaloutcomemeasurefor this review.Forthepatientsthemselves,however,theconsequencesof non-adherencemaybedifficulttograsp.Fromtheperspectiveof healthcarepersonnel,usingpharmaceuticalpictogramsmayserve toimprovethequalityoftheinformationprovidedtopatients,yet evaluatingtheimpactofpictogramsonpatientadherencemaystill prove difficult due to their indirect relationship. Other study endpointscouldrevealamoredirectrelationshipbetweentheuse ofpictogramsandclinicaloutcome,thoughforthetimebeing,we consideradherencetobethegoldstandard.

Funding

Thiswork was supportedin partbytheResearchCouncilof Norway and Norway Health Tech (FORREGION Oslo/Akershus

Bedriftsprosjekt–3027),andtheFacultyofNursingandHealth Sciences,NordUniversity,Norway.

DeclarationofCompetingInterest

H.S.andL.A.B.S.havenonetodeclare.E.A.T.isoneoftwoco- foundersofthecompany DepictASand hasdevelopedadigital pictogram-based tool for the provision of patient information regardingthecorrectuseofmedicines.

CRediTauthorshipcontributionstatement

Hege Sletvold: Conceptualization, Methodology, Validation, Formalanalysis,Investigation,Resources,Writing-originaldraft, Project administration, Funding acquisition. Lise Annie Bjørnli Sagmo:Conceptualization,Methodology,Validation,Formalanal- ysis,Investigation,Writing-review&editing.EirikA.Torheim:

Conceptualization, Methodology, Validation, Formal analysis, Investigation,Datacuration,Writing-review&editing,Visualiza- tion,Fundingacquisition.

References

[1]E.Sabaté,AdherencetoLong-termTherapies:EvidenceforActionAvailableat https://www.who.int/chp/knowledge/publications/adherence_report/en/, WorldHealthOrganization(WHO),Geneva,Switzerland,2003.

[2]P.Kardas,P.Lewek,M.Matyjaszczyk,Determinantsofpatientadherence:a reviewofsystematicreviews,Front.Pharmacol.4(2013)91,doi:http://dx.doi.

org/10.3389/fphar.2013.00091.

[3]R.L.Cutler,F.Fernandez-Llimos,M.Frommer,C.Benrimoj,V.Garcia-Cardenas, Economicimpactofmedicationnon-adherencebydiseasegroups:a systematicreview,Br.Med.J.Open8(2018)e016982,doi:http://dx.doi.org/

10.1136/bmjopen-2017-016982.

[4]H.P.Puspitasari, P.Aslani, I. Krass, Areview ofcounselling practices on prescriptionmedicinesincommunitypharmacies,Res.SocialAdm.Pharm.5 (2009)197–210,doi:http://dx.doi.org/10.1016/j.sapharm.2008.08.006.

[5]J.Grime,A.Blenkinsopp,D.K.Raynor,K.Pollock,P.Knapp,Theroleandvalueof writteninformationforpatientsaboutindividualmedicines:asystematic review,HealthExpect.10(2007)286–298,doi:http://dx.doi.org/10.1111/

j.1369-7625.2007.00454.x.

[6]G.A.Holt,J.D.Hollon,S.E.Hughes,R.Coyle,OTClabels:canconsumersreadand understandthem?Am.Pharm.NS30(1990)51–54.

[7]D.K.Raynor,A.Blenkinsopp,P.Knapp,J.Grime,D.J.Nicolson,K.Pollock,etal.,A systematicreviewofquantitativeandqualitativeresearchontheroleand effectivenessofwritteninformationavailabletopatientsaboutindividual medicines,HealthTechnol.Assess.11(2007)1–160.

[8]H.K.Chan,M.A.Hassali,C.J.Lim,F.Saleem,W.L.Tan,Usingpictogramstoassist caregiversinliquidmedicationadministration:asystematicreview,J.Clin.

Pharm.Ther.40(2015)266–272,doi:http://dx.doi.org/10.1111/jcpt.12272.

[9]I.M.Barros,T.S.Alcantara,A.R.Mesquita,A.C.Santos,F.P.Paixao,D.P.LyraJr., Theuseofpictogramsinthehealthcare:aliteraturereview,Res.SocialAdm.

Pharm.10(2014)704–719,doi:http://dx.doi.org/10.1016/j.

sapharm.2013.11.002.

[10]M.G. Katz, S.Kripalani, B.D.Weiss, Use of pictorial aids in medication instructions:areviewoftheliterature,Am.J.Health.Syst.Pharm.63(2006) 2391–2397,doi:http://dx.doi.org/10.2146/ajhp060162.

[11]M. Berthenet, R. Vaillancourt, A. Pouliot, Evaluation, modiﬁcation, and validationofpictogramsdepictingmedicationinstructionsintheelderly,J.

HealthCommun.21(Suppl.1)(2016)27–33,doi:http://dx.doi.org/10.1080/

10810730.2015.1133737.

[12]N.Kheir,A.Awaisu,A.Radoui,A.ElBadawi,L.Jean,R.Dowse,Developmentand evaluationofpictogramsonmedicationlabelsforpatientswithlimited literacyskillsinaculturallydiversemultiethnicpopulation,Res.SocialAdm.

Pharm.10(2014)720–730,doi:http://dx.doi.org/10.1016/j.

sapharm.2013.11.003.

[13]M.Montagne,Pharmaceuticalpictograms:amodel fordevelopment and testingforcomprehensionandutility,Res.SocialAdm.Pharm.9(2013)609–

620,doi:http://dx.doi.org/10.1016/j.sapharm.2013.04.003.

[14]D.Moher,A.Liberati,J.Tetzlaff,D.G.Altman,Preferredreportingitemsfor systematicreviewsandmeta-analyses:thePRISMAStatement,Syst.Rev.4 (2009)1,doi:http://dx.doi.org/10.1186/2046-4053-4-1.

[15]J.P.Higgins,D.G.Altman,P.C.Gotzsche,P.Juni,D.Moher,A.D.Oxman,etal.,The CochraneCollaboration’stoolforassessingriskofbiasinrandomisedtrials,Br.

Med.J.343(2011)d5928,doi:http://dx.doi.org/10.1136/bmj.d5928.

[16]J.A.Sterne,M.A.Hernan,B.C.Reeves,J.Savovic,N.D.Berkman,M.Viswanathan,etal., ROBINS-I:atoolforassessingriskofbiasinnon-randomisedstudiesofinterventions, Br.Med.J.355(2016)i4919,doi:http://dx.doi.org/10.1136/bmj.i4919.

[17]T.R.Vetter,M.E.Downing,S.C.Vanlandingham,K.M.Noles,A.M.Boudreaux, Predictorsofpatientmedicationcomplianceonthedayofsurgeryandthe