RESEARCHARTICLE Impact of a Daily SMS Medication Reminder System on Tuberculosis Treatment Outcomes: A Randomized Controlled Trial ShamaMohammed1*,RachelGlennerster2,AamirJ.Khan1 1 InteractiveResearchandDevelopment,Karachi,Pakistan,2 AbdulLatifJameelPovertyActionLab, MassachusettsInstituteofTechnology,Cambridge,MA,UnitedStatesofAmerica *[email protected] a11111 Abstract Importance Therapiduptakeofmobilephonesinlowandmiddle-incomecountriesoverthepast decadehasprovidedpublichealthprogramsunprecedentedaccesstopatients.Whilepro- OPENACCESS gramshaveusedtextmessagestoimprovemedicationadherence,therehavebeenno Citation:MohammedS,GlennersterR,KhanAJ high-poweredtrialsevaluatingtheirimpactontuberculosistreatmentoutcomes. (2016)ImpactofaDailySMSMedication ReminderSystemonTuberculosisTreatment Outcomes:ARandomizedControlledTrial.PLoS Objective ONE11(11):e0162944.doi:10.1371/journal. pone.0162944 TomeasuretheimpactofZindagiSMS,atwo-waySMSremindersystem,ontreatment Editor:LeiGao,ChineseAcademyofMedical successofpeoplewithdrug-sensitivetuberculosis. SciencesandPekingUnionMedicalCollege, CHINA Design Received:May9,2016 Weconductedatwo-arm,paralleldesign,effectivenessrandomizedcontrolledtrialinKara- Accepted:August26,2016 chi,Pakistan.IndividualparticipantswererandomizedtoeitherZindagiSMSorthecontrol Published:November1,2016 group.ZindagiSMSsentdailySMSreminderstoparticipantsandaskedthemtorespond Copyright:©2016Mohammedetal.Thisisan throughSMSormissed(unbilled)callsaftertakingtheirmedication.Non-respondents openaccessarticledistributedunderthetermsof weresentuptothreeremindersaday. theCreativeCommonsAttributionLicense,which permitsunrestricteduse,distribution,and reproductioninanymedium,providedtheoriginal Setting authorandsourcearecredited. PublicandprivatesectortuberculosisclinicsinKarachi,Pakistan. DataAvailabilityStatement:Datacanbeaccessed athttps://dataverse.harvard.edu/dataset.xhtml? Participants persistentId=doi:10.7910/DVN/HVJ2CM. Funding:Thisstudywasfundedbythe Newly-diagnosedpatientswithsmearorbacteriologicallypositivepulmonarytuberculosis InternationalInitiativeforImpactEvaluation(http:// whowereontreatmentforlessthantwoweeks;15yearsofageorolder;reportedhaving www.3ieimpact.org/),grantnumberOW1.48.The accesstoamobilephone;andintendedtoliveinKarachithroughouttreatmentwereeligi- fundershadnoroleinstudydesign,datacollection bletoparticipate.Weenrolled2,207participants,with1,110randomizedtoZindagiSMS andanalysis,decisiontopublish,orpreparationof themanuscript. and1,097tothecontrolgroup. PLOSONE|DOI:10.1371/journal.pone.0162944 November1,2016 1/13 ImpactofaDailySMSMedicationReminderSystemonTuberculosisTreatmentOutcomes CompetingInterests:Theauthorshavedeclared MainOutcome thatnocompetinginterestsexist. Theprimaryoutcomewasclinicallyrecordedtreatmentsuccessbaseduponintention-to- treat. Results WefoundnosignificantdifferencebetweentheZindagiSMSorcontrolgroupsfortreatment success(719or83%vs.903or83%,respectively,p=0(cid:1)782).Therewasnosignificantpro- grameffectonself-reportedmedicationadherencereportedduringunannouncedvisitsdur- ingtreatment. Conclusion Inthislarge-scalerandomizedcontrolledeffectivenesstrialofSMSmedicationreminders fortuberculosistreatment,wefoundnosignificantimpact. TrialRegistration ThetrialwasregisteredwithClinicalTrials.gov,NCT01690754. Introduction Tuberculosisisthesecond-leadingcauseofdeathfrominfectiousdiseasesglobally,withnine millionpeopleinfectedand1.5milliondeathsin2013[1].Treatment fordrug-sensitivetuber- culosislastssixtoeightmonthsandcanresultindifficultsideeffects.Failuretoadheretotreat- mentcanresultincontinuedtransmission,thedevelopmentofmultidrug-resistant tuberculosis,ordeath.TheWorld HealthOrganizationrecommendsdirectlyobservedtherapy topromoteadherence,whereapre-assignedtreatmentsupporterwatcheseachpatienttake his/herdailymedication.However,evidencefortheeffectivenessofthismethodisinconclusive [2]. Therapiduptakeofmobilephonesinlow-andmiddle-incomecountrieshasprovidedpub- lic-healthprogramsunprecedentedaccesstopatients.Mobilephone-basedinterventionsto improvemedicationadherencehavebeenadoptedformanydiseases,withmixedresults[3–6]. Mosttrialsareinadequatelydesigned,insufficientlypowered,orarerestrictedtohigh-income countries[3–6].Themostrigoroustrialsevaluatingmobilephone-usefortreatmentcompli- anceindevelopingcountriesexistforadherencetoantiretroviraltherapy(ART)forpeopleliv- ingwithHumanImmunodeficiencyVirus(HIV).Meta-analysisthatcombinestwopositive trialsinKenya[7,8]andonenullresultfromCameroon[9]foundanoverallpositiveimpactof weeklyShortMessageService(SMS,ortextmessage)reminders[10–12].Amorerecentstudy ofinteractiveautomatedvoiceremindersandpictorialmessagesinIndia,foundnoimpacton adherencetoART[13]. Whilethereisconsiderableinterestinthepotentialofmobilephone-basedinterventionsto improvetuberculosistreatmentadherence,rigorousevidenceontheirimpactislimited[14– 17].Recently,alimitedclusterrandomizedtrialinChinafoundthat,whilemedicationmonitor remindersledtoimproveddrugcompliancefortuberculosispatients,dailytwo-waySMS remindersdidnot[18].Oureffectivenesstrialgaugedtheimpactofatwo-wayinteractiveSMS medicationremindersystem(ZindagiSMS)onthetreatmentsuccessofpeoplewithdrug-sen- sitivetuberculosis. PLOSONE|DOI:10.1371/journal.pone.0162944 November1,2016 2/13 ImpactofaDailySMSMedicationReminderSystemonTuberculosisTreatmentOutcomes Methods StudyDesignandParticipants Participantswererecruitedintotherandomizedcontrolledtrialthroughalargetertiarycenter (theIndusHospital),ninepublicfacilities,andanetworkofprivateGeneralPractitioner(GP) clinicsandprivatelaboratoriesinKarachi. To beeligible,participantshadtobenewlydiagnosedwithsmearorbacteriologicallyposi- tivepulmonarytuberculosis;15yearsofageorolder;reporthavingaccesstoamobilephone; andintendtoliveinKarachithroughouttreatment.To allowZindagiSMStohelpestablish habitsearly,participantshadtobeontreatmentforlessthantwoweeks.To minimizespill- overs,patientswithanotherhouseholdmemberinthestudywereineligible.Enrolmentcontin- ueduntilthepredeterminedsamplesizewasmet. Eligibleparticipantswereconsentedusinganoralconsentform,acopyofwhichwasgiven totheparticipant.Oralconsentwassolicitedinsteadofwrittenconsent,asoursamplewasa low-literatepopulation.Thetrialandoralconsentprocedurewasapprovedbytheresearcheth- icsboardsatInteractiveResearchandDevelopment(IRD)inKarachi,PakistanandtheMassa- chusettsInstituteofTechnology(MIT)inCambridge,USA.Thetrialwasregisteredwith ClinicalTrials.gov,NCT01690754. RandomizationandMasking Onceapatientconsentedtoparticipate,astudyrepresentativeenteredidentifyinginformation onamobilephone-basedenrolmentform.Individualparticipantswererandomizedtoeither theZindagiSMSorcontrolgroups,usingpredeterminedlistonthestudyserverthatwasgen- eratedusingsimplerandomization.Theresearchteamwasblindedtotheallocationsequence generated.Ifmobiledataconnectivitywasinterrupted(10%ofenrolments),thestudyrepre- sentativecalledtheirsupervisor,whoenteredtheidentifyinginformationintoMicrosoftExcel andgeneratedthegroupassignmentusingtherandomizationfunction.Inbothcasesfieldstaff whointeracteddirectlywithpatientshadnoabilitytoinfluencetherandomizedallocation. Therandomizationstatusofindividualparticipantswasnotsharedwithtreatingclinicsbythe researchteam. Procedures Allstudyparticipantsreceivedthestandardofcareprovidedbytheirclinic.Initially,partici- pantsreceivedNTP’s recommendedeight-monthtreatmentregimen.However,NTPguide- lineschangedinthesecondquarterof2012toasix-monthregimen. ZindagiSMSusedtwo-wayreminderstoencouragepatientstoactivelyengagewithremind- ers,ratherthanpassivelyreadandpotentiallyignorethem.Itenabledthestudyteamtoidentify non-responsivepatientsforphonecallstoencourageparticipation.ZindagiSMSsentenrolled patientsdailyremindersatatimeoftheirchoosing.MessageswereinUrduusingEnglish scriptandincludedadailymotivationalmessagefollowedbyaremindertorespondviaSMS or,afterSeptember2011,amissed(unbilled)calltoindicatetheyhadtakentheirmedication. Fourteenmessageswererandomizedandsenttoparticipants.Basedonfeedbackfromour pilot,tuberculosiswasnotmentionedinthemessagesduetostigma[19].Forexample,one messagesaid,“Your healthisinyourhands.Take yourmedicationandremembertorespondby SMSoramissedcall.”SMSresponseswerenotverifiedforcontent.Participantswereoffered PKR60(USD0.60)permonthtocoverthecostsofresponding.Initially,theparticipantswere askedtopickupreimbursementsattheirclinic,butfromOctober2013,reimbursementswere transferreddirectlytoparticipants’phones. PLOSONE|DOI:10.1371/journal.pone.0162944 November1,2016 3/13 ImpactofaDailySMSMedicationReminderSystemonTuberculosisTreatmentOutcomes OnceZindagiSMSreceivedaresponseoramissedcall,aconfirmatorySMSmessagewas senttotherespondent.Ifthepatientdidnotrespondwithintwohours,asecondreminderwas sent.Athirdandfinalreminderforthedaywassentaftertwoadditionalhoursofnon-respon- siveness.Membersofourstudyteamphonedparticipantswhodidnotrespondforsevendays. Participantswereinterviewedattheirhouseholdforamoreextensivebaseline.Ninetyper- centofcompletedbaselinesurveyswereconductedwithin30daysofenrolment.Oncea month,studyenumeratorsattemptedsurprisevisitstoparticipants’householdstoconducta midlinesurvey.Sputumsamples,independentfromthosecollectedbythetreatingclinics,were collectedforpatientsatvariouspointsintheirtreatment.Giventhedifficultiesinfindingpar- ticipantsathomeandthattheycouldnotalwaysproducesputum,thereisconsiderablevaria- tioninthesputumsamplesavailableperpatient.Anendlinesurveywasconductedafterthe completionofparticipants’treatmentperiod.Participantswhowerereportedashaving defaultedortransferredoutfromtreatmentweresurveyedagainbetweenSeptemberand November2014torecordwhethertheycontinuedtheirtreatment. Theenrolmentformwasenteredonmobilephones,withpaperformsasbackup.Othersur- veyswerecollectedonpaperanddouble-enteredintoaMicrosoftAccessdatabase. Outcomes Theprimaryoutcomewasprogrammaticallydefinedtreatmentsuccessasrecordedinclinic registersprovidedbytheNTP, theglobalstandardforassessingtuberculosisprograms.Treat- mentsuccessisdefinedasthesumofpatientsclinicallyreportedascured(i.e.patientwhose sputumsmearorculturewaspositiveatthebeginningoftreatmentbutwhowassmear-orcul- ture-negativeinthelastmonthoftreatmentandonatleastonepreviousoccasion)ortreat- mentcompleted(i.e.apatientwhocompletedtreatmentbutwhodoesnothaveanegative sputumsmearorcultureresultinthelastmonthortreatmentandonatleastoneprevious occasion).Inoursecondaryanalysis,weexploredthefullrangeofclinicallyrecordedtreatment outcomes(cured,treatmentcompleted,default,died,treatmentfailureandtransferredout).In particular,ZindagiSMSmighthaveincreasedsuccessbyreducingdefault(i.e.apatientwhose treatmentwasinterruptedfortwoconsecutivemonthsormore).Thelimitationofthisout- comeisthatbecauseregistersofNTP-reportingclinicsinPakistanarenotlinked,patientswho arereportedashavingdefaultortransferredoutcouldhavecontinuedtreatmentelsewhere.As arobustnesscheckwethereforeexploredadjustedtreatmentoutcomesfortreatmentsuccess anddefault,substitutingself-reportedforclinicallyrecordedoutcomesfortheparticipants interviewed. To explorepotentialmechanismsofimpact(orlackofimpact),wealsogaugedadherence asasecondaryoutcomebyaskingparticipantswhethertheyhadtakentheirmedicationinthe last24hoursduringhomevisits.Whileself-reportedadherencecanbeunreliable,thereisno reasontobelievemisreportingissystematicallydifferentamongthoseassignedtoZindagiSMS orthecontrolgroup.DuringsurveyvisitsbetweenFebruaryandApril2012,weconductedIso- Screentests,whichdetectisoniazidmetabolitesinurinesamplestogaugewhethertuberculosis drugsweretakenwithinthepast24hours.Isoniazidisalwaysincludedinfirst-linetuberculosis treatment.We comparedIsoScreenresultswithself-reportedadherenceonthesamevisit. WhileIsoScreentestswereconductedonanon-randomsample,theygiveusanindicationof thereliabilityofself-reportedadherence. Finally,wecollectedself-reportedpsychologicalandphysicalhealthmeasuresassecondary outcomes.We usedafour-pointLikertscaleforpatients’difficultyincompletingarangeof physicaltasksandhowsupportedtheyfeltduringtreatment,apictureofafive-rungladder withsixcompartmentsbetweenrungsforlikelihoodofbeingcured,andimagesoffivefaces PLOSONE|DOI:10.1371/journal.pone.0162944 November1,2016 4/13 ImpactofaDailySMSMedicationReminderSystemonTuberculosisTreatmentOutcomes forhowhealthytheyfelt.Questionsonparticipants’abilitytocompletetasksandself-reported healthwereaskedatbaseline,midline,andendline;thelikelihoodofbeingcuredaskedatbase- lineandmidline;andhowsupportedparticipantsfeltaskedatendline. StatisticalAnalysis We calculatedaminimumsamplesizeof1,094participantsineachstudyarm,usingpowerof 80%,minimumdetectableeffectsize(MDE)offivepercentagepoints,andtreatmentsuccess rateinthecontrolgroupof75%.AnMDEof5percentagepointswaschosenbecauseasmaller impactwasconsideredunlikelytomotivatepolicychange.Datawereanalyzedusingintention- to-treat,i.e.allocationtotreatmentnotactualtakeup.We usedtheχ2testfordifferencesin proportionsfortheanalysisoftheprimaryoutcome.Incomparingsputumsamples,weran ordinaryleastsquaresregressions,aftercontrollingfordaysinthestudyandregimentype. We alsoassessedwhetherZindagiSMShaddifferentialimpactsontreatmentsuccessby gender,indicatorsofqualityofcare,andmobilephoneaccess.Inanalyzingoursecondaryout- comes,weranordinaryleastsquaresregressionsoneachoutcometotestforprogrameffects, aftercontrollingfordayssinceenrollmentandregimentype.To adjustformultiplehypotheses testingweusedtheBonferronicorrectionandthelessconservativeWestfallandYoung free step-downresamplingmethod[20]. StatisticalanalysiswasconductingusingSTATA/IC version12.0.Ap-valueof<0.05was consideredstatisticallysignificant. Results We enrolled2,207participantsintothestudybetweenMarch18,2011andFebruary25,2014 (Fig1).Bothgroupshadsimilarbaselinecharacteristics(Table 1). Asaneffectivenesstrialwesoughttoreplicateimplementationconditions,asitwouldexist atscale.TheZindagiSMSsystemsentremindersorreceivedresponsesfor174,284patientdays toparticipantsduringtheirestimatedtreatmentduration(180daysforthesix-monthregimen or240daysfortheeight-monthregimen).Withperfectimplementation,remindersshould havebeensentforatotalof220,560patientdays,suggestingthesystemwassuccessfullyimple- mentedfor79%ofpatientdays.Missedreminderswereduetosystemfailures,administrative shortfalls,orGPRSoutagesinthecitymobile(14%);participantsaskingtoleavethesystemor dying(3%);participantsoptingoutofreceivingremindersatenrolment(2%);andparticipants notknowingtheirphonenumberatenrolmentandfailingtosharetheirnumbersubsequently (2%).Ofthe1,069participantswhoweresentmessages,912(85%)respondedatleastonce.Of theparticipantsthatwereonthesystemthroughouttheirtreatment,themeanresponserate (calculatedasthenumberofresponsesreceivedoverthenumberofreminderssentperpatient) was29%,rangingfrom0to99%.Overthecourseoftreatment,averageresponseratesfellfrom 48%inthefirsttwoweeksto24%(eight-monthregimen)and20%(six-monthregimen)inthe lasttwoweeks(SeeFig2). Therewerenosignificantdifferencesintreatmentoutcomesbetweenparticipantsinthe ZindagiSMSgroupandthecontrolgroup(Table 2).Asarobustnesscheck,wesubstitutedthe clinicallyrecordedoutcomeswithself-reportedoutcomesforparticipantswhodefaultedor transferredoutoftreatmentthatwewereabletointerview,orwhosefamilymembersreported theparticipanthaddied.Ofthe283participantsthatwerereportedashavingdefaultedor transferredout,weinterviewed130(46%);49participants(17%)haddied;22(8%)refusedto beinterviewed;andwewereunabletolocate82(29%).Self-reportedoutcomeswerecatego- rizedusingthecriteriainFig3.Whenweadjustedtreatmentoutcomestoreflecttheself- reportedoutcomes,thedefaultratereducedinbothgroupsbuttherewerestillnosignificant PLOSONE|DOI:10.1371/journal.pone.0162944 November1,2016 5/13 ImpactofaDailySMSMedicationReminderSystemonTuberculosisTreatmentOutcomes Fig1.TrialProfile. doi:10.1371/journal.pone.0162944.g001 differencesinoutcomesbetweenthetwogroups(Table 2).We alsoexaminedindependent sputumsamplesbetweenthegroups.We collected1191sputumsamplesforparticipants(603 intheZindagiSMSgroupand588inthecontrolgroup).Ofthese,onehundredandninety sampleswereexcludedbecausetheywereprimarilysaliva,ratherthansputum(104intheZin- dagiSMSgroupand86inthecontrolgroup).Therewasnostatisticallysignificantdifference betweensputumresultsforbothgroups(p=0.762). Therewasnosignificantprogrameffectinsubgroupsforgender,anyoftheindicatorsrelat- ingtoqualityofcare,oraccesstoamobilephoneafteradjustingformultiplesubgroupsusing theBonferronicorrection(adjustedp-valueof0.003for17subgroups)ortheWestfalland Young freestep-downresamplingcorrection(Table 3).We alsocreatedasingleindexof PLOSONE|DOI:10.1371/journal.pone.0162944 November1,2016 6/13 ImpactofaDailySMSMedicationReminderSystemonTuberculosisTreatmentOutcomes Table1. Demographiccharacteristicsofparticipantsenrolledinthetrial(2011–2014,Karachi,Paki- stan).* SMSGroup(n=1110) ControlGroup(n=1097) n(%) n(%) Female 561(51%) 518(47%) Age(mean/SD)† 33(16) 33(16) Urduismothertongue 529(48%) 549(50%) Clinictype IndusHospital 404(36%) 385(35%) GPclinic/private‡laboratory 190(17%) 193(18%) Publictuberculosisclinic 516(46%) 519(47%) 6Monthtreatmentregimen 764(69%) 777(71%) Assignedatreatmentsupporter 102(10%) 106(10%) Ownmobilephone 540(49%) 565(52%) Schooling Noschool 517(49%) 475(47%) Primary(class1–5) 108(10%) 115(11%) Secondary(class6–10) 325(31%) 307(30%) Tertiary(aboveclass10) 77(7%) 101(10%) Religiousschool 15(1%) 16(2%) *Thereare138missingvaluesforassignedatreatmentsupporter,145missingvaluesforage;146missing valuesfornoschoolandreligiousschool;and151missingvaluesforprimary,secondary,andtertiary. †SD=standarddeviation ‡GP=privategeneralpractitioner doi:10.1371/journal.pone.0162944.t001 qualityofcarefromrelevantvariablesandanotherformobilephoneaccesstotestforsubgroup effectbutagainfoundnosignificanttreatmenteffectsforsubgroups.(S1Appendix) We comparedtheresultsof159IsoScreentestswithself-reportedadherenceduringthe samestudyvisit.IsoScreentestsindicatedthat17%ofthosewhosaidtheyhadtakentheir drugsinthepast24hourshadnot,indicatingover-reporting.However,theseresultswerenot statisticallydifferentbytreatmentarm. Therewerenosignificantprogrameffectsinself-reportedmedicationadherenceoranyof ourothersecondaryoutcomes,afteradjustingformultiplehypothesesusingtheBonferroni correction(p-valueof0.001forfivehypotheses)ortheWestfallandYoung correction (Table 4). Discussion We foundnosignificantimpactofSMSmedicationremindersontreatmentsuccess,other treatmentoutcomes,self-reportedadherence,orself-reportedphysicalandpsychological health.Therewerenostatisticallysignificantimpactsofremindersontreatmentsuccesswithin avarietyofsubgroupsafteradjustingformultiplehypothesistesting. Withanattritionrateoflessthan1%,similaracrossarms,onourprimaryoutcomevariable, theresultsareveryrobust.Theinclusionofparticipantsfrompublicclinics,privateGPclinics, laboratories,andalargephilanthropichospitalcontributestoitsexternalvalidity. Alimitationofourtrialwasthelackofanobjectiveadherencemeasure.Adherencewas self-reported,which,IsoScreentestsindicated,isanoverestimateofactualadherence.How- ever,evenifparticipationinZindagiSMSgeneratedincreasedadherence,notreflectedinself- reports,itdidnottranslateintoimprovedtreatmentoutcomes.Inaddition,IsoScreentests(on PLOSONE|DOI:10.1371/journal.pone.0162944 November1,2016 7/13 ImpactofaDailySMSMedicationReminderSystemonTuberculosisTreatmentOutcomes Fig2.Responseratesovertimeintreatment. doi:10.1371/journal.pone.0162944.g002 anonrandomsample)indicatedthatmisreportingwassimilarintheZindagiSMSandcontrol groups. Anotherpotentiallimitationisthatclinicscouldincorrectlyrecordtreatmentoutcomesto meetexpectedsuccessratesencouragedbytheNTP. However,sinceclinicswereblindtoallo- cation,thereisnoreasonanymisreportingwassystematicallydifferentbetweentreatment arms.Moreover,inoursub-studyonparticipantsreportedashavingdefaultedortransferred out,self-reportedoutcomesweresimilarintheinterventionandcontrolgroups.Another Table2. Clinically-recordedtreatmentsuccessbetweenZindagiSMSandcontrolgroups(2011–2014,Karachi,Pakistan). Clinicallyrecordedtreatmentsuccess Clinicallyrecordedtreatmentsuccessadjustedforself- reportedoutcomes ZindagiSMS ControlGroup ZindagiSMS ControlGroup n % N % p-value n % n % p-value Treatmentsuccess 917 83% 903 83% 0.782 923 84% 911 83% 0.871 Treatmentcomplete 332 30% 325 30% 0.863 339 31% 333 30% 0.903 Cured 585 53% 578 53% 0.960 584 53% 578 53% 0.994 Default 108 10% 103 9% 0.775 74 7% 80 7% 0.572 Died 19 2% 19 2% 0.975 38 3% 29 3% 0.282 TreatmentFailure 27 2% 29 3% 0.758 26 2% 29 3% 0.655 TransferOut 33 3% 39 4% 0.446 43 4% 44 4% 0.875 Total 1104 1093 1104 1093 doi:10.1371/journal.pone.0162944.t002 PLOSONE|DOI:10.1371/journal.pone.0162944 November1,2016 8/13 ImpactofaDailySMSMedicationReminderSystemonTuberculosisTreatmentOutcomes Fig3.Criteriafordeterminingself-reportedoutcomes. doi:10.1371/journal.pone.0162944.g003 potentiallimitationisthatourtrialtookplaceinPakistan,whichreportshightreatmentsuc- cessratesof91%in2012[1].Resultsmaydifferincountriesreportinglowersuccessrates. OurresultsweresimilartothefindingsofLiuandcolleaguesinChinathattwo-waySMS remindershadnoimpactonmedicationadherencefortuberculosispatients,thoughweare abletoruleoutmuchsmallereffectsofSMSreminders[15].Meta-analysisofstudiesonHIV andSMSremindersshowpositiveresultsonofweeklySMSremindersonadherenceandclini- caloutcomes(Finitsisetal2014[10],Mbuagbawetal2013[11],Horvathetal2012)[12]. Thesereviewsdrawmainlyonthreelargehighqualitystudies(Finitsisetalalsoincludemany smallpoorqualitystudiesandHorvathetalonlyincludesLesterandcolleagues’[7]andPop- Elechesandcolleagues’[8]trials)andconcludeweeklyremindersaremoreeffectivethandaily reminders.ThethreekeystudiesareMbuagbawandcolleagues,whotestweeklySMSremind- ersinCameroonandfindnoimpact[9]Pop-Elechesandcolleagueswhotestedmultiplever- sionsofSMSremindersandonlyfoundonlyshortmessagessentweeklyimproveadherence [8],andLesterandcolleaguesfindapositiveimpactofweeklyremindersinKenya[7].Impor- tantly,inLesterandcolleagues,participantsalsoreceivedoff-siteclinicianfollow-upifthey PLOSONE|DOI:10.1371/journal.pone.0162944 November1,2016 9/13 ImpactofaDailySMSMedicationReminderSystemonTuberculosisTreatmentOutcomes Table3. Sub-groupanalysisusingtreatmentsuccessastheoutcome(2011–2014,Karachi,Pakistan). ZindagiSMS Control ZindagiSMS naive FWER‡-adjusted Group Group Sub-group N % n % coefficient p-value* p-value Sex Male 438 80% 468 81% -0.012 0.618 0.999 Female 479 86% 435 84% 0.019 0.394 0.974 Qualityofcare IndusHospital 317 79% 301 78% 0.005 0.872 1 GPClinic/PrivateLab 180 95% 167 87% 0.082 0.006† 0.069 PublicTBClinic 420 82% 435 84% -0.023 0.331 0.954 Assignedatreatmentsupporter 89 87% 84 79% 0.080 0.124 0.698 Notassignedatreatmentsupporter 793 84% 768 84% -0.004 0.815 1 Remindedtotakemedication(withonemonthafterenrolment) 355 82% 324 83% -0.011 0.689 1 Notremindedtotakemedication(withinonemonthafterenrolment) 436 87% 446 84% 0.022 0.318 0.953 Accesstoamobilephone Ownmobilephone 450 84% 474 84% -0.005 0.806 1 Don’townmobilephone 467 82% 429 81% 0.016 0.502 0.994 Noschooling 416 81% 381 80% 0.004 0.875 1 Anyschooling 461 88% 469 87% 0.006 0.758 1 Atleastoneliteratepersoninthehousehold 753 85% 744 84% 0.008 0.629 0.999 Noliteratepeopleinthehousehold 129 80% 108 81% -0.016 0.736 1 CansendSMS(withinmonthofenrolment) 263 88% 232 87% 0.014 0.624 0.999 CannotsendSMS(withinfirstmonthofenrolment) 528 83% 538 82% 0.001 0.958 1 *Bonferronicorrectionp-value(with17subgroups):0.003 †p<0.05 ‡Family-wiseerrorrate20 doi:10.1371/journal.pone.0162944.t003 reportedaproblemordidnotrespondtotheSMS[7,21].NotincludedinthereviewsisShet andcolleagueswhotestweeklyautomatedtelephoneremindersinIndiaandfindnosignificant effect[13]. Table4. SecondaryoutcomesbetweentheZindagiSMSandcontrolgroups(2011–2014,KarachiPakistan). Tookmedication Perceptionsonlikelihoodof Howhealthytheyfelt† Easeofcompleting Howmuchsupportwas inthelast24 beingcured†(6=verylikely, (5=veryhealthy, tasks†(4=nodifficulty, received†(4=lotof hours† 1=notlikely) 1=veryunhealthy) 1=lotofdifficulty) support,1=nosupport) Zindagi 0.002 -0.008 -0.012 -0.017 0.020 Naïvep- 0.772 0.473 0.423 0.036‡ 0.521 value* FWER§ 0.89 0.89 0.89 0.162 0.89 adjustedp- value N(surveys) 11,301 9,560 11,324 11,235 1658 N(patients) 2091 2068 2091 2088 1658 *Bonferronicorrectionp-value(with5hypotheses):0.001 †Controllingforthelengthoftheregimen,daysinthestudy,anddaysinthestudy-squared. ‡p<0.05 §Family-wiseerrorrate[20] doi:10.1371/journal.pone.0162944.t004 PLOSONE|DOI:10.1371/journal.pone.0162944 November1,2016 10/13
Description: