ebook img

Improving the quality of child anthropometry PDF

13 Pages·2017·1.7 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Improving the quality of child anthropometry

RESEARCHARTICLE Improving the quality of child anthropometry: Manual anthropometry in the Body Imaging for Nutritional Assessment Study (BINA) JoelConkle1*,UshaRamakrishnan1,2☯,RafaelFlores-Ayala1,3☯,Parminder S.Suchdev1,2,3,4☯,ReynaldoMartorell1,2☯ 1 DoctoralPrograminNutritionandHealthSciences,LaneyGraduateSchool,EmoryUniversity,Atlanta,GA, UnitedStatesofAmerica,2 HubertDepartmentofGlobalHealth,RollinsSchoolofPublicHealth,Emory University,Atlanta,GA,UnitedStatesofAmerica,3 DivisionofNutrition,PhysicalActivityandObesity; a1111111111 NationalCenterforChronicDiseasePreventionandHealthPromotion,U.S.CentersforDiseaseControland a1111111111 Prevention,Atlanta,GA,UnitedStatesofAmerica,4 DepartmentofPediatrics,SchoolofMedicine,Emory a1111111111 University.Atlanta,GA,UnitedStatesofAmerica a1111111111 a1111111111 ☯Theseauthorscontributedequallytothiswork. *[email protected],[email protected] Abstract OPENACCESS Anthropometricdatacollectedinclinicsandsurveysareofteninaccurateandunreliabledue Citation:ConkleJ,RamakrishnanU,Flores-Ayala R,SuchdevPS,MartorellR(2017)Improvingthe tomeasurementerror.TheBodyImagingforNutritionalAssessmentStudy(BINA)evalu- qualityofchildanthropometry:Manual atedtheabilityof3Dimagingtocorrectlymeasurestature,headcircumference(HC)and anthropometryintheBodyImagingforNutritional armcircumference(MUAC)forchildrenunderfiveyearsofage.Thispaperdescribesthe AssessmentStudy(BINA).PLoSONE12(12): protocolforandthequalityofmanualanthropometricmeasurementsinBINA,astudycon- e0189332.https://doi.org/10.1371/journal. pone.0189332 ductedin2016–17inAtlanta,USA.Qualitywasevaluatedbyexaminingdigitpreference, biologicalplausibilityofz-scores,z-scorestandarddeviations,andreliability.Wecalculated Editor:FasilTekolaAyele,NationalInstitutesof Health,UNITEDSTATES z-scoresandanalyzedplausibilitybasedonthe2006WHOChildGrowthStandards(CGS). Forreliability,wecalculatedintra-andinter-observerTechnicalErrorofMeasurement Received:June28,2017 (TEM)andIntraclassCorrelationCoefficient(ICC).Wefoundlowdigitpreference;99.6% Accepted:November22,2017 ofz-scoreswerebiologicallyplausible,withz-scorestandarddeviationsrangingfrom0.92 Published:December14,2017 to1.07.TotalTEMwas0.40forstature,0.28forHC,and0.25forMUACincentimeters. Copyright:Thisisanopenaccessarticle,freeofall ICCrangedfrom0.99to1.00.ThequalityofmanualmeasurementsinBINAwashighand copyright,andmaybefreelyreproduced, similartothatoftheanthropometricdatausedtodeveloptheWHOCGS.Weattributedhigh distributed,transmitted,modified,builtupon,or qualitytovigoroustraining,motivatedandcompetentfieldstaff,reductionofnon-measure- otherwiseusedbyanyoneforanylawfulpurpose. TheworkismadeavailableundertheCreative menterrorthroughtheuseoftechnology,andreductionofmeasurementerrorthroughade- CommonsCC0publicdomaindedication. quatemonitoringandsupervision.Ouranthropometrymeasurementprotocol,whichbuilds DataAvailabilityStatement:Dataandthestudy onandimprovesupontheprotocolusedfortheWHOCGS,canbeusedtoimproveanthro- measurementmanualareavailableatOpen pometricdataquality.Thediscussionillustratestheneedtostandardizeanthropometric ScienceFramework(identifiers:DOI10.17605/ dataqualityassessment,andweconcludethatBINAcanprovideavaluableevaluationof OSF.IO/WMBS2|ARKc7605/osf.io/wmbs2). 3Dimagingforchildanthropometrybecausethereiscomparisontogold-standard,manual Funding:TheBillandMelindaGatesFoundation measurements. (www.gatesfoundation.org)fundedthestudy(ID OPP1132308).Dr.ReynaldoMartorellwasthe PrincipalInvestigatorforthestudy.Allauthorsdo nothaveaffiliationswithorfinancialinvolvement withanyorganizationorentitywithafinancial PLOSONE|https://doi.org/10.1371/journal.pone.0189332 December14,2017 1/13 BINAmanualanthropometrymethodsandquality interestinthesubjectmatterormaterials Introduction discussedinthemanuscript.Fundingsourceshad TheMulticenterGrowthReferenceStudy(MGRS)andsubsequentdevelopmentofWorld noroleindataanalyses,datainterpretation,or reportwriting.Thecorrespondingauthorhadfull HealthOrganization(WHO)ChildGrowthStandards(CGS)in2006providedasinglesetof accesstoalldata,finalresponsibilityforthe referencemeasurementsforchildrenaroundtheglobe[1].TheWHOCGShavebeencom- decisiontosubmitforpublication,andhasno monlyadaptedforroutineuseinlow-andmiddle-incomecountriesaswellasinsomehigh- conflictsofinteresttodisclose.Thefindingsand incomecountries,includingtheUS,forpurposesofindividualgrowthmonitoring,clinical conclusionsinthisarticlearethoseoftheauthors research,andpublichealthprogrammonitoring[2].Withthenewreferencecameasingle anddonotnecessarilyrepresenttheofficial measuringprotocolthatcouldbeadoptedinhealthfacilitiesandsurveyseverywhere[3];how- positionoftheCentersforDiseaseControland Prevention. ever,theMGRSprotocolrequiresextensivetrainingandsupervision,andrepeatedmeasure- ments[3].ThefullMGRSprotocolisnotroutinelyfollowedinhealthfacilities,norisitused Competinginterests:Theauthorshavedeclared inlarge-scalesurveysthatevaluatenutritionsuchasDemographicandHealthSurveys(DHS) thatnocompetinginterestsexist. andMultipleIndicatorClusterSurveys(MICS). Thenon-digital,manualmeasurementscurrentlyinusearesusceptibletohumanerror[4], andtheuseofinadequatemeasuringprotocolscanincreasemeasurementerror.Poorquality childanthropometryiscommoninbothhealthfacilitiesandsurveys,withskinfoldthickness, circumferencesandchildlengththemostaffected[5–9].Measurementerrorcancausemis- classificationofnutritionalstatusattheindividuallevelandoverdispersionatthepopulation level;thelatterleadstooverestimationoftheprevalenceofabnormalnutritionalstatus. Thereisrecognitionthatthequalityofchildanthropometryshouldbeassessedbefore usinganddisseminatingthedata.DHS,MICSandStandardizedMonitoringandAssess- mentofReliefandTransitions(SMART)Surveymethodologiesincludeassessmentof anthropometricdataquality.Theassessmentsarealllooselybasedonrecommendations fromaWHOexpertcommitteeconvenedin1995[10],buttherearemethodologicaldiffer- ences.Recently,therehavebeencallstoimproveanthropometryqualitythroughtheuseof technology,andtorevisitthe1995WHOrecommendationstostandardizedataquality assessment[11]. ThedataweanalyzeispartoftheBodyImagingforNutritionalAssessmentStudy(BINA), whichcompareda3Dimagingsystemtocurrentlyrecommendednon-digital,manualmea- surementsofstature(lengthandheight),armcircumference(MUAC)andheadcircumference (HC)in474childrenunderfiveyearsofageinAtlanta,USA.BINAusedtheAutoAnthro3D imagingsystem,asystemcomprisedofanoff-the-shelf,low-cost,handheldscanner(Occipital Inc.StructureSensor,SanFrancisco,CA,USA)andcustomsoftwareforscanningchildren developedbyBodySurfaceTranslations,Inc.(BST)(AutoAnthro,BST,Inc.,Atlanta,GA, USA).Inordertodrawconclusionsontheabilityof3Dimagingtoreplicatemanualanthro- pometry,thestudyneededtocollectgold-standardmanualanthropometry.Athoroughanaly- sisofmanualanthropometryqualitywasparticularlyimportantforBINAbecause,unlike previousresearchon3Dimagingforanthropometry,thestudysampleconsistedofchildren underfiveyearsofage,anagegroupthatcanbedifficulttomeasurewithmanualmethods becauseofnoncooperation.Thispaperdescribesthetraining,standardizationanddatacollec- tionmethodsformanualanthropometryinBINA;evaluatesthequalityofBINAmanual anthropometry;andprovidessomerecommendationsforachievinggoldstandardmanual anthropometryandstandardizingdataqualityassessmenttoguideclinicians,researchersand publichealthprogrammanagers. Materialsandmethods PrimarycaregiversofallchildrenparticipatinginBINAgavewritten,informedconsentand thestudywasapprovedbytheEmoryInstitutionalReviewBoard. PLOSONE|https://doi.org/10.1371/journal.pone.0189332 December14,2017 2/13 BINAmanualanthropometrymethodsandquality Fieldstafftrainingandstandardization AllfivefieldstaffselectedfortheBINAStudyheldcollegedegrees,withthreeofthefivehold- ingamaster’sdegreeatthetimeofthestudy.InAugust2016fieldstaffcompletedathree- weektrainingledbytrainersfromEmoryUniversitywhohadextensiveexperiencewith anthropometryinclinic,survey,andresearchsettings;includingexperienceinthestudyused todevelopthe2006WHOCGS.Trainingconsistedoftheoreticalandpracticalsessionson3D imagingandmanualmeasurements,andfieldstaffweretrainedtofunctionasbothanthropo- metristsandassistants.EmoryUniversityfacultyandfieldstaffdevelopedtrainingmaterials andastudymanual. Trainingculminatedwithathree-daystandardizationtestformanualanthropometryata localdaycarecenter,whichconsistedofaleadanthropometristfromEmoryUniversityandall fieldstafftakingrepeatedmeasurementsoftenchildrenunderfiveyearsofage.Thelead anthropometristwasconsideredanexpertanthropometristbasedonpreviousexperienceand areliabilitytestcarriedoutbeforeBINA.Asanthropometrists,fieldstaffwereassessedonall measurementsexceptweight,anddatawasanalyzedusingENASoftware2011[12]anda MicrosoftExcelSpreadsheetfromtheUnitedStatesCentersforDiseaseControlandPreven- tion(USCDC)MicronutrientSurveyToolkit[13].Passingorfailingthestandardizationtest wasbasedonaccuracyandreliabilityresultsforthemainmeasurementsofinterest(length andheight).Wedeterminedaccuracybycomparingeachanthropometristtotheleadanthro- pometristandtothemeanofallanthropometrists.Forreliability,wecomputedtheTechnical ErrorofMeasurement(TEM),whichdeterminesifanthropometristsgetsimilarresultswhen carryingoutrepeatedmeasurements.WealsousedtheUSCDCspreadsheetstovisuallypres- entaccuracyandreliabilityresultstotheanthropometrists,andtoassesswhetherornotan anthropometrist’sfirstmeasurementwassystematicallylowerorhigherthanthesecondmea- surement,alsoknownasthemeasurementeffect[3].Therewasnoevidenceofsubstantial measurementeffect.Theaccuracyandreliabilityoftheleadanthropometristandallanthropo- metristsforlengthandheightweresimilartoresultsfromtheMGRS[3]andwereclassifiedas thehighestrankingof“good”accordingtotheSMARTsuggestedcut-offpointof<0.4cmfor bothintra-observerTEMandbiasfromanexpert[14].Basedonresultsofthestandardization test,weretainedallanthropometristsforthestudy. Sampling,measurementanddataentryprocedures Utilizingconveniencesamplingwerecruitedandmeasuredchildreninfacilities,primarily daycarecenters,inAtlanta,GA,USAfromSeptember2016toFebruary2017.Formanual anthropometrywefollowedmeasurementtechniquesusedtodevelopthe2006WHOGrowth Standards[3],includingmeasuringchildrenundertwoyearsofagelyingdownandmeasuring tothenearesttenthofakgandcm.Wemeasuredweightwithdigitalscaleswithtaringfunc- tion(RiceLakeWeighingSystems,Inc.,RiceLake,WI),staturewithaninfant/child/adult woodenboard(ShorrBoard,WeighandMeasure,LLC,MarylandUSA),andcircumferences withsyntheticmeasuringtapes(ShorrTape,WeighandMeasure,LLC,MarylandUSA).We routinelycheckedcalibrationofscalesusingknownweightsandreplaceddamagedmeasuring tapes. Duringdatacollectionsessions,thefivefieldstaffsplitintotwoteamsoftwo,withthefifth personactingasafloater.Thetechniquesforscansandmanualmeasurementsrequiredan assistant,andteammatesalternatedastheanthropometristandassistant.Anthropometrists alsotookturnsasafloater,andthemainrolesofthefloaterweretopreparechildrenformea- surementandtoactasasecondassistantforyoungerchildren.Anthropometryprocedures designedforhouseholdsurveysoftencontainaroleforthechild’sparent.However,sincewe PLOSONE|https://doi.org/10.1371/journal.pone.0189332 December14,2017 3/13 BINAmanualanthropometrymethodsandquality obtaineddataprimarilyindaycarecenterswherethecaregiverswerenotpresent,asecond assistantwasneededtoholdandpositionthechild. Aftertakingtimetoacclimatethechildandestablishrapport,andafterundressingthe child(totheirdiaperortoskin-tightshorts/leotardsprovidedbythestudy),thefieldteam startedwith3Dscanswhichwasthenfollowedbymanualmeasurements.Thefirstanthropo- metristcompletedonesession,measuringheadcircumference,MUAC,andlengthorheight. Insteadofoneanthropometristcompletingtwosessionsconcurrently,thefieldteamreversed rolesandanewanthropometristcompletedherfirstsessionofmanualmeasurements.After bothteammemberscompletedtheirfirstsession,theprocesswasrepeatedforthesecondses- sion.Westaggeredmanualmeasurementstoreducethelikelihoodofanthropometrists rememberingtheirfirstmeasurement.Tofurtherminimizebiasininter-observererror, anthropometristsenteredtheirownmeasurementresultsanddidnotinformtheassistantof thenumber,whichisdifferentfromthetypicalanthropometryprocedureoftheassistant recordingresultsonbehalfoftheanthropometrist. Qualitycontrolanddatacleaning Ourcustomsoftwareforelectronicdatacaptureincludedrangechecksfornon-digitalmea- surementstocatchdataentryerrors.Ifameasurementwasbelowthe0.01percentileorabove the99.9percentile,apop-upboxappearedonthescreenandtheanthropometristcouldeither re-enteroracceptthevalue.Thesoftwarealsoincludedautomatictriggeringofathirdmea- surementfornon-digitalmeasurements.ThetriggerswereprogrammedbasedontheMGRS standardsforthemaximumallowabledifference((cid:20)0.5cmforheadandarmcircumference, and(cid:20)0.7cmforlength)[3],butourstudydifferedfromMGRSinthatwetriggeredwithin observerdifferenceswhileMGRStriggeredbetweenobservers.Foreachchild,twoanthropo- metristsentereddemographicinformation,includingdateofbirth,separately.Doubledata entryallowedustoidentifydiscrepanciescausedbydataentryerror,andwereferredtothe originalconsentformtomakecorrections. Formanualanthropometry,fieldstaffreceivedongoingmonitoringandsupervision.This includedexaminingdigitpreference,thepercentofintra-observermeasurementsexceeding themaximumallowabledifference,inter-observerTEM,andestimatesofbiasbycomparing toanexpertanthropometristwhomeasuredasubsampleofchildren.Inadditiontocomparing toanexpertanthropometrist,fieldstaffalsoperiodicallycomparedwitheachother.Fieldstaff regularlyreceiveddataqualityreportscoveringallmeasuresofqualityandsupervisionfrom expertanthropometrists. Qualitytests Fordigitpreference,weexaminedtheproportionofnon-digital,manualmeasurements (height,headcircumferenceandarmcircumference)witheachofthepossibledigits(0–9)in thetenthsplace(mm).Weanalyzedthefirstmeasurementfromeachanthropometrist.Since eachchildwasmeasuredbytwodifferentanthropometrists,weincluded948observationsin theanalysis.WeusedtheStataSE13(StataCorp,CollegeStation,TX,USA)digdispackagefor analysis.WetestedforanydigitpreferencewithPearson’sChiSquaredTest.Wecalculated thedifferencebetweentheexpectedandobservedpercentagesandtestedforpreferenceof eachindividualdigitusingatwo-sidedBinomialTest.Todeterminehowcloseourobserved proportionsweretoauniformdistribution,wesummedthedifferencebetweenexpectedand observedforpositivedifferencesonly.Thesumgivesthepercentageofdigitsthatwouldneed tobereclassified(movedfromanoverrepresenteddigittoanunderrepresenteddigit)to achieveauniformdistribution,whichissimilartotheMyer’sBlendedIndex[7]andtheDigit PLOSONE|https://doi.org/10.1371/journal.pone.0189332 December14,2017 4/13 BINAmanualanthropometrymethodsandquality PreferenceScore[14].WecalculatedtheDigitPreferenceScoretoallowcomparisonswith otherresearch. Weusedthe2006WHOGrowthStandardstocalculatez-scoresforweight-for-length/ height(WHZ),length/height-for-age(HAZ),weight-for-age(WAZ),headcircumference-for- age(HCZ),andarmcircumference-for-age(ACZ).Weassessedtheplausibilityofz-scoresby determiningtheproportionofmeasurementsflaggedasfallingoutsideoftheplausiblerange asdefinedinWHOmacros(WHZ<-5|>5;HAZ<-6|>6;WAZ<-6|>5;HCZ<-5|>5;ACZ <-5|>5)[15].WealsoexaminedflagsusedinDemographicandHealthSurveys(DHS)forthe rangeofplausiblelengthandheightmeasurements.IntheWHOMacro,staturemeasure- mentsoutsideoftheDHSplausiblerange(lyingdown45–110cm,standingup65–120cm) arenotflagged,butWHZscorescannotbecalculatedandareautomaticallysettomissing.We calculatedz-scorestandarddeviations(SD)andanalyzedSDdisaggregatedbyage(underand overtwoyearsofage)todetermineifthequalityofmeasurementsdifferedbyage.Wealso analyzedz-scoresforboththeaverageofallmeasurements(repeated)andthefirstmeasure- mentofthefirstanthropometrist(single). Weanalyzedbothintra-andinter-observererrorforheight,HCandMUACusingrepeated measurements.Intra-observertechnicalerrorofmeasurement(TEM)wascalculatedwiththe followingformula: rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P TEMintra¼ N ðM (cid:0) M Þ2= i¼1 i1 i2 2(cid:3)N ,whereNisthenumberofchildrenandM M aretheclosestrepeatedmanualmeasuresfor i1 i2 onechildbyoneobserver.Forinter-observerTEMwecomparedtheaveragemeasurementof oneobservertotheaveragemeasurementofanotherobserverforthesamechild,changingthe numeratorintheequationabove: rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P TEMinter ¼ N ðM þM =2(cid:0) M þM =2Þ2= i¼1 ij1 ij2 ik1 ik2 2(cid:3)N WecalculatedRelativeTEM,alsoknownas%TEM,bydividingTEMbythemeanofallthe measurementsthatwentintocalculatingTEMandmultiplyingby100[16].TotalTEMcom- binesintraandinter-observerreliabilityinthefollowingformula: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi TotalTEM¼ TEMintra2þTEMinter2 Forcorrelation,wecalculatedtheIntraclassCorrelationCoefficient(ICC)usingtwo-way mixedeffects[17]andanabsoluteagreementdefinitioninSPSS20(IBMCorp.,Armonk,NY, USA). Results SamplecharacteristicsinTable1showthatchildrenundertwoyearsofagewereoverrepre- sented,andthattherewassubstantialsize,racial,andethnicvariationinthesample. Qualitytests Digitpreference. Allmeasurements(stature,HC,andMUAC)showedevidenceoftermi- naldigitpreferencewhenwetestedthechild’sfirstmeasurementwithPearson’sChi-Squared Test(p<.01,n=948).Forallmeasurements,theterminaldigitfourwassignificantlyoverrep- resented.Inaddition,eightwasoverrepresentedforHC,andsixwasoverrepresentedfor MUAC(Table2).Thesumofthedifferencebetweenobservedandexpectedpercentagesfor PLOSONE|https://doi.org/10.1371/journal.pone.0189332 December14,2017 5/13 BINAmanualanthropometrymethodsandquality Table1. Samplecharacteristics. Ageinmonths,mean(range) 25.7 (0–59) AgeGroups Newborn(<1month) 82 (17%) 1–11.9months 66 (14%) 1year 75 (16%) 2years 85 (18%) 3–4.9years 166 (35%) Sex Female 228 (48%) Race Black 201 (42%) White 134 (28%) Asian 40 (8%) Multiple,OtherorNotReported 99 (21%) Ethnicity Non-Hispanic 385 (81%) Hispanic 77 (16%) NotReported 12 (3%) Statureincm,mean(range) 82.3 (42.5–119.0) HeadCircumferenceincm,mean(range) 45.7 (31.4–55.3) ArmCircumferenceincm,mean(range) 15.4 (8.8–25.3) https://doi.org/10.1371/journal.pone.0189332.t001 alloverrepresenteddigitsindicatedthat5.8%ofstaturemeasurementsand9.7%ofbothhead andarmcircumferencemeasurementswouldneedtobereclassifiedtoachieveauniformdis- tribution(Table2).Thedigitpreferencescorewas5.5forheight,8.7forHC,and8.3for MUAC. Plausibility&reliability. Usingmeansofrepeatedmeasuresandcutoffsdefinedby WHOforbiologicallyplausiblez-scoresonechild(0.2%)wasflagged,fallingabovetherange forWHZ,andACZ.Also,thelengthofonechild(0.2%)waslowerthantheplausiblestature range.Withsinglemeasures,thelengthofoneadditionalchildwaslowerthantheplausible staturerange,andnoadditionalchildrenhadimplausiblez-scores. Table2. Terminaldigitpreferenceexpressedaspercentageofanthropometrists’firstmeasurementforheight,headcircumferenceandarmcir- cumferenceendingin.0to.9comparedtotheexpected10%amongchildren0–4.9years(n=948observationsfrom474children),BINA2017. TerminalDigit Height HeadCircumference ArmCircumference (tenthsplace) % %Observed—% P-Value % %Observed—% P-Value % %Observed—% P-Value Observed Expected Observed Expected Observed Expected 0 8.5 -1.5 0.14 5.0 -5.0 0.00 5.9 -4.1 0.00 1 9.2 -0.8 0.45 9.2 -0.8 0.45 10.7 0.7 0.48 2 11.2 1.2 0.23 11.2 1.2 0.23 8.4 -1.6 0.12 3 10.2 0.2 0.79 10.1 0.1 0.87 9.8 -0.2 0.91 4 14.3 4.3 0.00 14.3 4.3 0.00 14.6 4.6 0.00 5 9.7 -0.3 0.83 6.8 -3.2 0.00 6.4 -3.6 0.00 6 9.8 -0.2 0.91 9.8 -0.2 0.91 12.3 2.3 0.02 7 9.8 -0.2 0.91 9.6 -0.4 0.75 11.6 1.6 0.10 8 9.1 -0.9 0.36 13.3 3.3 0.00 9.7 -0.3 0.83 9 8.1 -1.9 0.06 10.8 0.8 0.42 10.5 0.5 0.55 https://doi.org/10.1371/journal.pone.0189332.t002 PLOSONE|https://doi.org/10.1371/journal.pone.0189332 December14,2017 6/13 BINAmanualanthropometrymethodsandquality Formeansofrepeatedmeasures,thestandarddeviationforallz-scores(HAZ,WHZ, WAZ,ACZ,andHCZ)wascloseto1.0,rangingfrom0.92forWHZto1.07forHAZ.Standard deviationsusingsinglemeasureswereslightlyhigher,withthelargestdifferencesinACZ (0.03)andWHZ(0.02)(Table3).Withincreasedvariationwealsofoundsinglemeasureshad slightlyhigherprevalenceestimatesofchildrenbeloworabovesomeSDcutoffsforACZand WHZ(Fig1),butdifferenceswerenotstatisticallysignificantwhentestedwithChi-Square. Forbothrepeatedandsinglemeasures,z-scorestandarddeviationwasnotconsistentlyhigher forchildrenunder2yearsofagecomparedtochildren2yearsandolder.Levene’sTestfor EqualityofVariancesshowednostatisticaldifferenceinz-scorevariancebetweentheage groupsforallindices(Table3),indicatingthattherewasnodifferenceinmeasurementreli- abilitybyage. Thecorrelationbetweenananthropometrist’sfirstandsecondmeasurementofachild (intra-observer),andbetweentwoanthropometrists’averagemeasurementsofachild(inter- observer)wasnearperfect.TheIntraclassCorrelationCoefficientforallintra-andinter- observermeasurementswasexactly1.00(95%confidenceintervalof1.00,1.00)exceptfor inter-observerarmcircumference,withICCof0.99(CI:0.99,0.99).Intra-observerTEMin centimeterswas0.22,0.13,and0.16forstature,HC,andMUACrespectively.Intra-observer TEMvaluescorrespondedtorelativeTEMsof0.26%,0.29%,and1.04%respectively(Fig2). AsanillustrationofTEMinterpretation,thestatureintra-observerTEMof0.22cmmeans that2/3rdsofrepeatedmeasurementswerewithin±0.22cmand95%ofreplicatemeasure- mentswerewithin2(cid:3)TEM,or±0.44cm.Inter-observerTEMwashigherthanintra-observer forallmeasurements:statureTEM0.34cm,%TEM0.42%;HCTEM0.25cm,%TEM0.54%; andMUACTEM0.19cm,%TEM1.22%(Fig2).AlthoughMUACinterandintraTEMwas lowerinabsoluteterms,MUACrelativeTEMwasapproximatelytwotothreetimeshigher thanstatureandHC.TotalTEM,combininginter-andintra-observerreliability,was0.40for stature,0.28forHC,and0.25forMUACincentimeters. Table3. Meanz-score,standarddeviation,andtestofequalvariancebetweenagegroupsforheight-for-age(HAZ),weight-for-height(WHZ), weight-for-age(WAZ),armcircumference-for-age(ACZ),andheadcircumference-for-age(HCZ)frommanualmeasurements(n=474),BINA2017. Total(0–4.9years) Lessthan2years(U2) 2–4.9years(O2) DifferenceinVariance (Levene’sTest) N Meanz- Standard n Meanz- Standard n Meanz- Standard U2 F P-Value score Deviation score Deviation score Deviation SD-O2 (SD) SD Repeated Height-for-age 474 -0.29 1.07 223 -0.42 1.10 251 -0.18 1.03 0.08 0.12 0.73 measure Weight-for-height 472 0.34 0.92 222 0.32 0.88 250 0.35 0.96 -0.08 0.87 0.35 mean Weight-for-age 474 0.06 1.04 223 -0.05 1.02 251 0.17 1.05 -0.04 0.26 0.61 Arm 385 0.78 0.94 135 0.84 0.92 250 0.75 0.95 -0.03 <0.01 0.93 circumference- for-age Head 474 0.24 1.02 223 0.11 1.01 251 0.35 1.01 0.00 0.32 0.57 circumference- for-age Single Height-for-age 474 -0.30 1.08 223 -0.43 1.11 251 -0.17 1.02 0.09 <0.01 0.96 measure Weight-for-height 471 0.34 0.95 221 0.34 0.92 250 0.34 0.97 -0.05 0.38 0.54 Weight-for-age 474 0.06 1.04 223 -0.05 1.01 251 0.17 1.05 -0.04 0.31 0.58 Arm 385 0.78 0.97 135 0.85 0.98 250 0.75 0.96 0.02 0.25 0.62 circumference- for-age Head 474 0.24 1.04 223 0.11 1.04 251 0.36 1.03 0.01 0.52 0.47 circumference- for-age https://doi.org/10.1371/journal.pone.0189332.t003 PLOSONE|https://doi.org/10.1371/journal.pone.0189332 December14,2017 7/13 BINAmanualanthropometrymethodsandquality Fig1.Classificationcomparison.Prevalencebystandarddeviationcutoffsforweight-for-heightandarmcircumference-for-agez-scoresforsingle andmeansofrepeatedmeasuresamongchildren0–4.9yearsofage. https://doi.org/10.1371/journal.pone.0189332.g001 Discussion OurdatashowsthatmanualanthropometryinBINAwasofexcellentquality.Only0.4%ofthe samplehadbiologicallyimplausiblez-scoresormeasurements,whichiswellbelowthe1.0% cutoffrecommendedbyaWHOexpertcommitteeasanindicatorofdataqualityproblems [10].Ourz-scoreSDsalsoindicatedgoodqualitybecausetheywerebetween0.9SD—1.1SD [14,18],andtherewerenodifferencesinSDbetweenagegroups;ahigherSDamongchildren under2isanindicatorofpoorqualityandisattributedtodifficultyinmeasuringthelengthof young,uncooperativechildren[7,19].Wefoundnoterminaldigitpreferenceforzeroorfive forstature,andthepercentofmeasurementstheoreticallyrequiringreclassificationforanor- maldistributionwaslowforallmeasures.Wecouldobtainanormaldistributionbychanging 6%ofterminaldigitsforourstaturemeasurementscomparedtoanaverageof18%in52DHS from2005–2014[7].Biologicalplausibility,z-scoreSD,anddigitpreferencearemetricscom- monlyusedtoassessdataqualityofsingleanthropometricmeasures[3,7–9],andourstudy showedexcellentqualityaccordingtoallthreemetrics.Inaddition,therepeatedmeasuresin ourstudyenabledanalysisofinter-andintra-observerreliability.ForstaturetheBINAintra- andinter-observerTEMsof0.22and0.34cmrespectivelyarewithintherangeofinter-observer TEMatMGRSstudysites(0.15–0.41cm)[4].TheTEMsforallofourmeasurementswereon parwithTEMsobservedatMGRSstudysitesandwerewithinthe95%precisionmarginof MGRSexpertanthropometrists[4].ThequalityofmanualanthropometryinBINAwassimilar tothequalityofanthropometricdatausedtodevelopthe2006WHOgrowthstandards. Inourstudy,weachievedhighqualitymanualanthropometrythroughfollowingestab- lishedprotocolsandaddingadditionalqualitycontrolcomponents.Wefollowedadvicefrom PLOSONE|https://doi.org/10.1371/journal.pone.0189332 December14,2017 8/13 BINAmanualanthropometrymethodsandquality Fig2.Measurementreliability.Closesttwomanualmeasuresfromsingleobserver(intra-observer)andmeasurementreliabilityinaverageof closesttwomanualmeasuresbetweentwoobservers(inter-observer)forstature,headcircumference(HC)andarmcircumference(MUAC)among children0–4.9yearsofage(intra-observern=948,inter-observern=474). https://doi.org/10.1371/journal.pone.0189332.g002 MGRSprotocolauthorstodevelopastudyspecifictrainingmanual,trainandteststaffwith standardization,andprovideadequatesupervisionduringdatacollection[3].Formeasure- menttechniques,wedrewfrommaterialsusedinhouseholdsurveys[20–23],whicharelargely basedona1986UNpublicationonmeasuringchildren[24].Wealsomadesurethatourtech- niqueswerealignedwiththeproceduresusedtodeveloptheWHOChildGrowthStandards byreferringtoMGRSprotocolandtheWHOanthropometryvideousedtotrainMGRS anthropometrists[3,25].Withtheaimtostrengthentheirunderstandingof,anddedicationto adhereto,studyprotocols,weincludedourfieldstaffindevelopmentofthestudymanual. Fieldstaffproducedthefirstdraftofthe3Dimagingsectionofthemanualandparticipatedin revisionandrefinementofmanualanthropometrysections.Fieldstaffparticipationinstudy manualdevelopmentledtorecognitionoftheneedtoimproveuponsomeoftheborrowed materials.Forexample,theinstructionsforfindingthemid-upperarmpointdidnotade- quatelyexplainidentificationoftheacromionprocessandillustrationsincorrectlydepicted measuringthesideofthearm.Forheadcircumference,weaddedadditionalinstructionsfor theanthropometriststorepositionatboththefrontandbackoftheheadwhileremovingand replacingthetapemultipletimestofindthelargestcircumference.Electronicdatacapture allowedustoavoiddataentryerrorsthroughtheuseofrangechecksanddoubledataentryat thepointofdataacquisition;andfacilitatedadequatemonitoringandsupervision.Identifying errorsduringmeasurementallowedustore-measurechildren,reducingthenumberofim- plausiblemeasurements.LiketheMGRSandfollowingrecommendationsfroma1995expert committee[3,10],weroutinelyassessedfieldstaffaccuracyandprovidedregular,timelyfeed- backonthequalityofanthropometry.Assessmentofaccuracyduringdatacollectionisnot standardincommonsurveyssuchasDHSandMICS.Repeatedmeasurements,whicharerare outsideofresearchstudysettingsallowedustoincludereliabilityinqualitymonitoring;and PLOSONE|https://doi.org/10.1371/journal.pone.0189332 December14,2017 9/13 BINAmanualanthropometrymethodsandquality electronicdatacaptureprovidedanthropometristswithreal-timefeedbackontheirownreli- abilitythroughtheuseofautomatictriggers. Therehavebeensuggestionstotakerepeatedmeasuresinsurveys[11].Ourresultsshowlit- tledifferencebetweensingleandrepeatedmeasuresforz-scorestandarddeviationsandpreva- lence.However,withpoorqualityanthropometryrepeatedmeasuresshouldreducevariance. Also,triggeringadditionalmeasurementsprovidesanincentivetoanthropometriststo improvereliabilityandavoidadditionalwork.Forsurveyinstitutionswithaknownhistoryof poorqualityanthropometrytemporaryadoptionofduplicatemeasuresbytwoanthropome- trists,alongwithtriggeringofrepeatedmeasuresbasedoninter-observermaximumallowable difference,mayhelptoimprovequalityandprovideaccurateresults.Inter-observererrorwas higherinourstudy,whichisconsistentwithfindingsinadevelopingcountrysetting[19];the MGRSusedinter-observertriggeringandadoptionofinter-observertriggersinsurveysmay improvequalitymorethanintra-observertriggers.Inaddition,repeatedmeasurescouldbe restrictedtothemostunreliablemeasures,suchaslengthofchildrenundertwoyearsofage andMUAC. Thereissomeconsensusonwhattoassessforanthropometricdataquality.A1995WHO expertcommitteerecommendedlookingataccuracy,reliability,biologicalplausibility,z-score standarddeviation,anddigitpreference[10];andthesemetricsareincorporatedintomanuals andassessmentsforDHSandMICS.SMARTsurveymethodologyincludesallofthesemet- rics,alongwithadditionalmetricssuchasskewnessandkurtosis.The2011EmergencyNutri- tionSoftware(ENA)forSMARThasautomatedreportsforstandardizationtestsanddata quality,withthelatterreferredtoasaplausibilitycheck[14].Whilequalityassessmentsfor DHS,MICSandSMARTaresimilar,themethodsarenotexactlythesame.Forexample,z- scorerangesforbiologicalplausibilitycandifferandtherearemultipletestsfordigitprefer- ence.Ourstudyprovidesasimplifiedmethodforassessingdigitpreferencethatdoesnot requirespecializedsoftware.Standardizationofwhattomeasureandhowtomeasureanthro- pometricdataqualityisneeded,buttheremaybelargerinstitutionaldifferencesinhowto interpretandactondataqualityreports.Oneofthechallengesinthisstudywasdetermining thecriteriatojudgewhetherornotweachievedgoldstandardmanualanthropometry.DHS andMICSdonothaveindicatorthresholdsforacceptableanthropometricdataquality,andit isveryrareforeithertosuppressdatabecauseofpoorquality.TheENASMARTplausibility checkdeterminesifdataqualityisacceptablewithacompositescorebasedonthresholdsfor multipleindicators[14].AnothercompositescoringsystemwasdevelopedbyUNICEF-sup- portedresearch[9],butthesystemisnotregularlyused.Formanysurveysdataqualityis assessed,butthereisnostandardcriteriatojudgewhetherornotresultsshouldbereleased. Wecanconsiderz-scorestandarddeviationtoillustratetheimportanceofreachingconsen- susoninterpretationandaction.WHOandtheUSCDCpromotetheuseofnormativeranges ofSDtodetermineifsurveyqualityisacceptable[10,26],buttherangesarebasedonsurveys thathaveevidenceofpoordataquality[7,18].ThemostrecentDHSdataqualityassessment showedthat30of52countrieshadHAZSDgreaterthan1.5,butonlyonecountrysuppressed databecauseofpoorquality[7].AccordingtoSMARTdataqualityisnotacceptableifHAZ SDisabove1.2[14],andarecentmodelingstudyshowedthatSDof1.5canresultinsubstan- tialoverestimationofstuntingprevalence[18].Meanwhile,thepublishednormativerangefor HAZSDthatsomeorganizationsusetodeemdataqualityacceptableis1.35–1.95[10,26]. Digitpreferenceprovidesanotherexampleoftheneedtofocusoninterpretationandaction. Digitpreferenceisacommonphenomenoncausedbyhumantendencytoroundnumbers thathasbeenmonitoredforyearsasanindicatorofdataquality,butevenveryhighmeasure- mentdigitpreference(80%theoreticallyneedingreclassificationforanormaldistribution)has nomeaningfulimpactonz-scoremeansorprevalences[18].Weightandheightheapingmay PLOSONE|https://doi.org/10.1371/journal.pone.0189332 December14,2017 10/13

Description:
protocol for and the quality of manual anthropometric measurements in . Standards [3], including measuring children under two years of age lying
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.