Integrating rapid risk mapping and mobile phone call record data for strategic malaria elimination planning

MISSING IMAGE

Material Information

Title:
Integrating rapid risk mapping and mobile phone call record data for strategic malaria elimination planning
Series Title:
Malaria Journal
Physical Description:
Mixed Material
Creator:
Andrew J Tatem
Zhuojie Huang
Clothilde Narib
Udayan Kumar
Deepika Kandula
Deepa K Pindolia
David L Smith
Justin M Cohen
Bonita Graupe
Petrina Uusiku
Christopher Lourenço
Publisher:
Malaria Journal
Publication Date:

Notes

Abstract:
Background: As successful malaria control programmes re-orientate towards elimination, the identification of transmission foci, targeting of attack measures to high-risk areas and management of importation risk become high priorities. When resources are limited and transmission is varying seasonally, approaches that can rapidly prioritize areas for surveillance and control can be valuable, and the most appropriate attack measure for a particular location is likely to differ depending on whether it exports or imports malaria infections. Methods/Results: Here, using the example of Namibia, a method for targeting of interventions using surveillance data, satellite imagery, and mobile phone call records to support elimination planning is described. One year of aggregated movement patterns for over a million people across Namibia are analyzed, and linked with case-based risk maps built on satellite imagery. By combining case-data and movement, the way human population movements connect transmission risk areas is demonstrated. Communities that were strongly connected by relatively higher levels of movement were then identified, and net export and import of travellers and infection risks by region were quantified. These maps can aid the design of targeted interventions to maximally reduce the number of cases exported to other regions while employing appropriate interventions to manage risk in places that import them. Conclusions: The approaches presented can be rapidly updated and used to identify where active surveillance for both local and imported cases should be increased, which regions would benefit from coordinating efforts, and how spatially progressive elimination plans can be designed. With improvements in surveillance systems linked to improved diagnosis of malaria, detailed satellite imagery being readily available and mobile phone usage data continually being collected by network providers, the potential exists to make operational use of such valuable, complimentary and contemporary datasets on an ongoing basis in infectious disease control and elimination.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
All rights reserved by the source institution.
System ID:
AA00020089:00001

Full Text

PAGE 1

Integratingrapidriskmappingandmobilephone callrecorddataforstrategicmalariaelimination planning Tatem etal. Tatem etal.MalariaJournal 2014, 13 :52 http://www.malariajournal.com/content/13/1/52

PAGE 2

RESEARCHOpenAccessIntegratingrapidriskmappingandmobilephone callrecorddataforstrategicmalariaelimination planningAndrewJTatem1,2*,ZhuojieHuang3,4,ClothildeNarib5,UdayanKumar4,6,DeepikaKandula7,DeepaKPindolia3,4, DavidLSmith2,8,JustinMCohen7,BonitaGraupe9,PetrinaUusiku5andChristopherLoureno5,7AbstractBackground: Assuccessfulmalariacontrolprogrammesre-orie ntatetowardselimination,theidentificationof transmissionfoci,targetingofattackmeasurestohigh-riskareasandmanagementofimportationriskbecome highpriorities.Whenresourcesarelimitedandtransmi ssionisvaryingseasonally,approachesthatcanrapidly prioritizeareasforsurveillanceandcontrolcanbeva luable,andthemostappropriateattackmeasurefora particularlocationislikelytodifferdependingon whetheritexportsorimportsmalariainfections. Methods/Results: Here,usingtheexampleofNamibia,amethodforta rgetingofinterventionsusingsurveillancedata, satelliteimagery,andmobilephonecallrecordstosuppor teliminationplanningisdescribed.Oneyearofaggregated movementpatternsforoveramillionpeopleacrossNamibia areanalyzed,andlinkedwithcase-basedriskmapsbuilton satelliteimagery.Bycombiningcase-dataandmovement ,thewayhumanpopulationmovementsconnecttransmission riskareasisdemonstrated.Communitiesthatwerestrongly connectedbyrelativelyhigherlevelsofmovementwere thenidentified,andnetexportandimportoftravellersandinfectionrisksbyregionwerequantified.Thesemaps canaidthedesignoftargetedinterventionstomaxima llyreducethenumberofcasesexportedtootherregions whileemployingappropriateinterventionstomanageriskinplacesthatimportthem. Conclusions: Theapproachespresentedcanberapidlyupdatedand usedtoidentifywhereactivesurveillanceforboth localandimportedcasesshouldbeincreased,whichregions wouldbenefitfromcoordinatingefforts,andhowspatially progressiveeliminationplanscanbedesigned.Withimprovem entsinsurveillancesystemslinkedtoimproveddiagnosis ofmalaria,detailedsatelliteimagerybeingreadilyavailableandmobilephoneusagedatacontinuallybeingcollectedby networkproviders,thepotentialexiststomakeoperati onaluseofsuchvaluable,complimentaryandcontemporary datasetsonanongoingbasisininfectiousdiseasecontrolandelimination. Keywords: Humanmobility, Plasmodiumfalciparum malaria,Malariaelimination,Migration,Diseasemapping,Spatial analysis,Satelliteimagery,MobilephonesBackgroundSignificantprogressisbeingmadeinreducingthemorbidityandmortalityattributedtomalariaglobally[1-10], andtheGlobalMalariaActionPlan(GMAP)[11]articulatesalong-termvisionformalariaeradicationthrough shorter-termlocaleffortstoeliminatemalaria.Atotalof 36ofthe107malaria-endemiccountrieshavedeclared theyhaveanationalpolicyformalariaeliminationor arepursuingspatiallyprogressiveeliminationwithintheir borders[11-14]. Achievingeliminationrequiresare-orientationaway fromthesortsofuniversalpreventionandtreatment measuresthatdefineacontrolprogrammetowards targetedoperations,suchasidentifyingresidualtransmissionfoci,focusingvectorcontrolorparasite-basedattack measurestohigh-riskareas,identifyingandcuringboth asymptomaticandsymptomaticinfections,andmanaging *Correspondence: A.J.Tatem@soton.ac.uk1DepartmentofGeographyandEnvironment,UniversityofSouthampton, Southampton,UK2FogartyInternationalCenter,NationalInstitutesofHealth,Bethesda,MD, USA Fulllistofauthorinformationisavailableattheendofthearticle 2014Tatemetal.;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsoftheCreative CommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,and reproductioninanymedium,providedtheoriginalworkisproperlycited.TheCreativeCommonsPublicDomainDedication waiver(http://creativecommons.org/publicdomain/zero/1.0/)appliestothedatamadeavailableinthisarticle,unlessotherwise stated.Tatem etal.MalariaJournal 2014, 13 :52 http://www.malariajournal.com/content/13/1/52

PAGE 3

importationrisk[15].Manyoftheseoperationalrequirementscanbefacilitatedbyaccurateandtimelycreationof riskmaps.Suchmapscanhelpeliminationprogrammes understandtheepidemiologyofadisappearingdisease, andmayallowproactivedeploymentofvectorcontrol measurestohighriskareastopreventlocaltransmission andonwardspreadtootherreceptiveareas,orsuggest areaswhereactivecasedetectionmaybeusedtoidentify andtreatremainingparasitereservoirs[16].Parasiteratebasedmapsformalariahavenowbeenconstructed [17,18],butinfectionprevalenceisapoormetricformeasuringmalariaatverylowlevelsofendemicity(below5% parasiteprevalence)duetothelargesamplesizesurveys requiredforprecisemeasurementinsuchcontexts[19].In verylowtransmissionenviron ments,diagnosticallyconfirmedmalariaincidenceprovidesamoreusefulmeasure thanprevalence,andelimination-focusedprogrammesare buildingcapacitytorapidlyprovidesuchinformation, includingtheplaceofresidenceofcases[20].Suchasurveillancesystemisacrucialcomponentofanelimination strategy,butachievingandmaintainingeliminationwill requirefindingandcuringinfectionsthatmaybeasymptomaticormaynevercomeintocontactwithreporting healthfacilities[15].Suchinfectionscanbeidentified throughintensiveproactivesurveillance,butthegeneration ofcase-basedriskmapsathighspatialresolutionhasthe potentialtoremotelyidentifyregionsinwhichtransmissionislikelytobeoccurringmorequicklyandatsubstantiallylowercost. Riskmapsareessentialforknowingwheretoattack malaria,buttheyareinsufficientforastrategiceliminationplan.Attackingstrategicallyrequiresdeployingthe rightmeasuresintherightplaces,anddoingsoinaway thatgainsarenotlostduetomovementofpeopleand parasites.Forexample,anattempttoeliminatemalaria inHaitiinthe1960sthroughmassdrugadministration combinedwithDDT-sprayingfailedbecausethehighly mobilepopulationcontinuallyreintroducedparasites intoareasthathadjustbeencleared[21].Understanding humanmovement,whichcanprovideconnectionsbetweendisparatehigh-riskareas,iscriticaltodesigning appropriateeliminationstrategiesandavoidingresurgenceinpost-eliminationsettings[22,23].However,data onhumanmovementpatternsinmalaria-endemicregionshavebeendifficulttoobtain,andoftenrestrictedto localtravelhistorysurveysorcensus-derivedmigration data[22].Therapidglobalproliferationofmobilephones haspresentedunprecedentedopportunitiesformeasuringandunderstandinghumanmovementdynamics.The retrospectiveanalysisofbillionsofcalldetailrecords (CDRs),wherebytemporalsequencesofphonetowerlocationsthroughwhichusercommunicationswererouted areconvertedintomovementtrajectories[24-27],providinginformationonhumantravelforsamplesizesof millionsandatscalesofentirecountries.Previousstudies havedemonstratedthevalueofsuchdatawhencombinedwithparasiteprevalencemapsinprovidingquantitativeguidancetomalariaprogrammes[25,28,29],and mapping ‘ source ’ and ‘ sink ’ areasofnetinfectionexport orimport[30].However,ineliminationsettingswhere infectionprevalenceisaninappropriatemeasureand wherecase-basedmalariamapsareofgreaterutility,such approacheshaveyettobeapplied. Here,thepotentialofintegratingmobilephoneCDRs withrapidcase-basedmappinginprovidingadynamic evidencebasetosupportmalariaeliminationplanningin lowtransmissionsettingsisdemonstrated,usingNamibia asanexample.Between2004and2011,scaleupofvector controlandcasemanagementinterventionsinNamibia contributedtoaremarkabledeclineinreportedmalaria casesfrom610,800to14,400[31].Namibiaisrapidlyscalingupitsmalariaprogramme,withsignificantstrengtheningofitsdiagnosisandsurveillancesystemsplannedover thenextfiveyears,focusedonachievingeliminationby 2020.Whilethecountryhasaclearstrategicplanand recentlydraftednationaleliminationpolicyinplace[32], achievingitsgoalswillrequireaclearlydefinedstrategy todeployresourcestooptimaleffect.Theintegrationof movementdatawithcase-basedriskmapsforNamibia providesadynamicframew orkforunderstandingthe connectivitybetweenexistingandpotentialmalariarisk areasanddefining ‘ source ’ and ‘ sink ’ regions,whererelativelylargernumbersofparasitesmaybeexportedthan importedthroughtravel,and vice-versa .Targetingaggressiveattackmeasurestosourcecommunitieswill reducemalariabothattheirlocationsandthroughout thewiderregiontowhichitexportsparasites.Atthe sametime,sustainablemeasurestoreducereceptivityin sinkregionswillbeimportanttolimitonwardstransmissionfromimportedinfections.MethodsEthicalapprovalThisprojectwasapprovedbyEthicsandResearch GovernanceoftheUniversityofSouthampton(submission #7696).MappingmalariariskDe-identifieddataoncasesofmalariaconfirmedusing rapiddiagnostictests(RDTs)reportingtohealthfacilities acrossthethreehighesttransmissionregions,Kavango, OmusatiandCaprivi(Figures1,2,3and4)forthemalaria transmissionseasoninJanuarytoMay2011werecollected bytheNamibiaNationalVector-borneDiseasesControl Programme(NVDCP)inthecourseofroutinesurveillance.Thecommunityofresidenceofeachpatient,as reportedtonursesathealthfacilitiesatthetimeoftreatment,wasgeolocated.Atotalof109casesfrom74uniqueTatem etal.MalariaJournal 2014, 13 :52 Page2of15 http://www.malariajournal.com/content/13/1/52

PAGE 4

locationsinKavango,234casesfrom41uniquelocations inOmusatiand332casesfrom47uniquelocationsin Capriviweresuccessfullygeolocated.Theaverageageof casesacrosssettlementsanddistrictsshowednosystematic differencesorbiases.Thisindicatedthattransmissionwas likelynothighenoughinanylocationforsignificantimmunitytodevelopandresultinlowercaseloadsduetoimmunityeffects,ratherthanenvironmentaldrivers.The procedureforproducinghighresolutionriskmapsfrom thecaselocationdatafollowedcloselythatoutlinedin Cohen etal .[16]andisdescribedbelow.Furtherdetails areprovidedinAdditionalfile1. Spatialcovariatedatasetsrepresentingrainfall,temperature,elevation,temperaturesuitabilityfor Plasmodiumfalciparum ,topographicwetness,vegetation,land cover,distancetowater,infrastructure,andpopulation densityat250mresolutionwereassembled.Fulldescriptionsanddetailsofdatasetsourcesareprovidedin Additionalfile1.AstransmissioninNamibiaisstrongly seasonal,wherecovariatedatawereavailablebymonth, datafortheJanuary-Mayperiodwereusedtomatch peaktransmission,followingassessmentofmalariaseasonalityinNamibiafromaggregatedNVDCPsurveillancesystemdata(Figures5and6,Additionalfile1). Valuesforeachofthecovariateswereextractedforthe pointlocationsofcommunitieswithconfirmedcases and ‘ background ’ points,randomlyselectedfromacross populatedareasoftheregions,identifiedusingapopulationdensitydataset[33],withpointssampledonlyfrom gridcellswithpopulationestimatesof>0.1persons. Backgroundpointsdonotnecessarilyindicatetheabsenceoftransmission,butinsteadcharacterizetheenvironmentofthecountry[34]intheplaceswherepeople live.Travelhistoryinformationfrompatientswerenot available,thereforetoattempttocontrolforthefactthat patientsmayhaveobtainedinfectionsawayfromtheir Figure1 PredictedprobabilityofmalariacasesinJanuary-May2011forOmusatiregion. TheresidentiallocationofRDTconfirmedcases aremappedascrosses. Figure2 PredictedprobabilityofmalariacasesinJanuary-May2011forKavangoregion. TheresidentiallocationofRDTconfirmedcases aremappedascrosses. Tatem etal.MalariaJournal 2014, 13 :52 Page3of15 http://www.malariajournal.com/content/13/1/52

PAGE 5

placeofresidence,locationswith(i)justonecase,then (ii)withoneortwocases,weredropped,basedonthe assumptionthatmultiplecasesinalocationaremore likelytoberepresentativeoflocaltransmission,andthe outputresultscomparedtothemappingrunusingall casedatatoexaminehowsensitiveoutputsweretothe exclusionoftheseisolatedcases.Samplesof10,000 backgroundpoints[34,35]wereselectedforeachregion. FollowingCohen etal. [16],theregressiontreeclassificationapproach ‘ RandomForest ’ [36]wasapplied usingtheR[37]packageModelMaptomodeltheriskof casesoccurringineach250250mgridcell,bothseparatelywithineachofthethreeregionswherethecase dataoriginated,andcombinedtoundertakemapping acrossthewholeofnorthernNamibia.Regressiontrees createaseriesofrulestopartitionthedataintoasetof groupsthatareashomogenousaspossiblewithrespect totheoutcome[38].Forexample,onesuchrulemight differentiatethelocationsofcasehouseholdsfromthose ofcontrolhouseholdsbasedonelevationbelowacertainthreshold,whileanotherrulemightfurtherdivide thedatabasedonlevelsofvegetationwithinspecific bounds.IntheRandomForestapproach,thedataare repeatedlysplitaccordingtomanydifferentbranching ‘ trees ’ ofthistype,andthefinalpredictionismadeby averagingacrossalloftheindividualtrees[36].Toassess theaccuracyofmodelpredictions,80%oftheobserved caseswereselectedatrandomfortrainingthealgorithm, withtheother20%usedfortesting,withthisrepeated100 times.Allofthepredictorvariableswereincludedinthe fittingsteptoproduceamodelpredictingtheprobability ofcasesoccurringataparticularlocationasafunctionof thecombinedcovariates.Modelqualitywasassessedby examiningcalibrationplots[39],theareaunderthecurve (AUC)onreceiveroperatingcharacteristic(ROC)graphs andcorrelationstatistics[40].Thefittedmodelwasthen appliedinconjunctionwiththe250mspatialresolution griddeddatasetsofthepredictivevariablestogeneratea mapofpredictedhighseasoncaseriskacrossnorthern Namibia. Mobilephonecalldatarecords CDRscoveringthe12-monthperiodOctober2010to September2011wereprovidedbytheleadingmobile Figure3 PredictedprobabilityofmalariacasesinJanuary-May2011forCapriviregion. TheresidentiallocationofRDTconfirmedcasesare mappedascrosses. Figure4 PredictedprobabilityofmalariacasesinJanuary-May2011fornorthernNamibia(withnamedregionsmarked). Tatem etal.MalariaJournal 2014, 13 :52 Page4of15 http://www.malariajournal.com/content/13/1/52

PAGE 6

Figure5 HealthdistrictsofNamibia. Figure6 Plotshowingthetrendsintotalmonthlyphone-derivedmovementsandreportedmalariacasesinNamibia. Tatem etal.MalariaJournal 2014, 13 :52 Page5of15 http://www.malariajournal.com/content/13/1/52

PAGE 7

phoneserviceproviderinNamibia,MobileTelecommunicationsLimited(MTC),whoreported1.5millionsubscribersin2011,anda90%marketshare[41].Thedata wereobtainedthroughwrittenagreementsbetweenthe networkprovider,theNVDCP,andtheClintonHealth AccessInitiative(CHAI).Followingpreviousstudies [24-27,30],anonymizedrecordsaggregatedtothelevel ofcelltowerswereprovidedtoensurethatitwasimpossibletoidentifyindividuals. ForeachCDRfromacallortextmessage,thecaller andreceiver(identifiedusingananonymousID),the receivingtowerIDthatthecallwasroutedthrough,and thedateofthecallwererecorded.Acrossthe12-month period,atotalof9billioncommunicationsfrom1.19 millionuniqueSIMcardswereidentifiedinthedataset, representing85%oftheestimated1.4millionadult(aged over15yearsold)populationofNamibia[42].Recent dataonthetypicalagesofmobilephoneownersin NamibiawereobtainedfromtheUniversalServiceBaselineStudyoftheCommunicationsRegulatoryAuthority ofNamibia[41]andshowedthatwhilethemajorityof userswerebetween20and30yearsold,therewasa broadspreadacrossagegroups(Additionalfile2). Moreover,recentanalysessuggestthatsuchbiasesmay havealimitedeffectongeneralestimatesofhuman mobility[43]. Movementswithinurbanareaswerenotconsidered here,giventheprincipalfocusofthisstudyonexaminingregionalmovementpatterns.Therefore,phone towersandmovementsfallingwithintheboundariesof urbanextentsmappedusingtheGlobalRuralUrban MappingProjectUrbanExtent(GRUMP-UE)dataset [44],wereaggregatedsothatonlymovementsbetween differenturbanareasorbetweenruralandurbanareas remainedintheanalyseddataset.Thisreducedthedatasetoflocations,orphonecatchmentareas,from626to 402.Whileratesofcross-bordermovementscouldnot beascertainedfromthedata,duetothenetworkprovidersonlyoperatinganational-levelnetwork,those crossingovertheborderintoNamibiafromneighbouring countriescommonlyswitchtoalocalSIM-card(MTC, perscomm).Thismeantthatthemovementsofsuch travellersandmigrantswerecapturedinthedataset, althoughtheanonymizednatureoftheCDRsmeantthat theycouldnotbeidentified,northeirmovementsanalysedseparatelyfromNamibianresidents.Dailylocations werecalculatedforthesubscribersusingthelocationof callsandtextsatoneofthe402phonecatchmentareas acrossthecountry,followingmethodsoutlinedinother similarstudies[24-27,30].Subscriberswereassigneda catchmentareaastheir ‘ home ’ residencebywhere themajorityofnightswerespentthroughoutthefull 12-monthperiod.Movementsbetweenareaswerecalculatedbyexaminingthetemporalsequencesofcallsor textssent/receivedbysubscribersandassigningamovementtoanewareaandatimeofthismovewhenthe areathroughwhichtheircall/textwasroutedchanged. Further,ageneralmeasureofpopulationmobility,the ‘ radiusofgyration ’ [24]wascalculatedforcomparisonof mobilitydifferencesbetweenareas.Theradiusofgyration measuresthecharacteristicdistancetravelledbyauser overacertaintimeperiod(inthiscase,the12-month period),andhasbeenwidelyusedinotherCDR-based humanmobilitystudies[24,26,45]. Themobilephonedataprocessingoutlinedaboveenablesconstructionofaweightednetworkofmovements betweeneachphonecatchmentarea.Theidentification ofdistinctcommunitieswithinthisweightednetworkwas undertakenusingamodularityoptimizationalgorithm [46].Theapproachfindshighmodularitypartitionsof largenetworksandunfoldsacompletehierarchicalcommunitystructureforthenetwork.Insimplerterms,the approachidentifiesgroupsofareasthatareconnectedby highlevelsofmovementandcombinesthemintoasingle ‘ community ’ .Ratesofmovementwithincommunitiesare generallyhigherthanbetweenseparatecommunities.Such communitydetectionapproacheshavebeenusedinpreviousmalariastudiestoidentifycommunitiesofregionsthat areeitherstronglyconnectedbyhumanorparasitemovements,oraremoreisolated[47,48].Thecommunitydetectionalgorithmwasrunhereonthenetworksofhuman andcaseriskscaled(seebelow)movements,andthedifferencesexamined.PopulationandmalariaflowsandconnectivityMovementsofpeopleandtheirinfectionswereestimatedfortwotypesoftravellers,followingprevious approaches[25,29,30]:(i) ‘ Returningresidents ’ :Residents ofalocationwhovisitedariskareathenreturnedto theirhomelocation,potentiallybringinganinfection withthem,and(ii) ‘ Visitors ’ :Residentsofariskarea whovisitedanewlocationandpotentiallycarriedan infectionwiththem.Here,giventhemalariacasedata available,therelativestrengthsofconnectivitybetween locationsintermsofthecase-basedmalariariskswere examined,ratherthanattemptingtoestimateabsolute numbersofinfectionsmoving. Forreturningresidents,itwasassumedthattheriskof acquiringaninfectionattheirplaceofvisitisafunction ofthelevelofriskatthevisitedlocationandthelength ofstay[25,29].Therefore,asimplemetricofcumulative riskwascalculatedbyscalingthenumberofdaysspent atthevisitedlocationduringthemalariatransmission seasonmonths(January-May)bythemodelledriskvalue thereforeachreturningresidenttrip.Forvisitorstonew locationsduringthetransmissionseason,itwasassumed thattherelativeriskofeachvisitorcarryinganinfection canbequantifiedbytheestimatedlevelofriskattheirTatem etal.MalariaJournal 2014, 13 :52 Page6of15 http://www.malariajournal.com/content/13/1/52

PAGE 8

homelocations.Thesesimplemetricsdefinedimportationriskflownetworksforreturningresidents,visitors and,bycombiningthetwo,overallriskflow,which quantifiedtheconnectivit ythroughhumanmovement scaledbypredictedriskacrossnorthernNamibia. Throughoutthefocusisonflowsandconnectivitybetween locationsfortheJanuary-May2011period.Mapping ‘ sources ’ and ‘ sinks ’Throughrepeatedintroductionofmalaria,humanmovementcanmakeitappearthatanareaissustaining transmission.Targetingther elativelylargerexporter communities( ‘ sources ’ )ofinfectionswithaggressiveattackmeasuresislikelytohaveanimpactonthenumbers ofinfectionsbothatthatlocationandinsurrounding areasthatarenetimportersofinfections( ‘ sinks ’ ).Atthe sametime,sinkcommunitieswithsubstantialpotentialfor transmissionrepresentplaceswherereceptivity-lowering activities,suchasvectorcontrol,maybeimportantto managetheriskofimportedmalariaonanongoingbasis. Thissortofstrategicdeploymentofinterventionsislikely toincreasetheeffectoflimitedresources.Theestimation ofrelativemalariariskconnectivitymatricesdescribed aboveenabledidentificationofthenetexporters(sources) orimporters(sinks)perlocation.ResultsCase-basedmalariariskmappingUnivariateanalysesdemonstratedtheutilityofthemajorityofthespatialcovariatesindistinguishingcaselocationsfrom ‘ background ’ conditions(Additionalfile1). TheRandomForestmodelprovidedfurtherindication ofthisthroughstrongmodelpredictionperformance withAUC=0.96andcorrelation=0.82(Additionalfile1). Modelassessmentstestingdataandstratifiedbydistrictconfirmedtheaccuracyoftheapproachinits abilitytoidentifylocationsofcasesthatwerenotincludedatthetrainingstage(Additionalfile1).Judging bytherelativeinfluenceonthemodelpredictions, outputsweremostdependentuponthespatialcovariates thatquantifiedvegetationamounts,populationdensity, precipitation,andpresenceofwater.Leastimportantvariableswerethoserelatedtotemperature,elevationand remoteness.Resultsweresimilarwhenbrokendownby district(Additionalfile1),highlightingtheconsistencies inlikelyenvironmentaldriversoftransmissionacross northernNamibia.Moreover,resultsappearedinsensitivetodroppinglocationswithonlyoneortwocases (Additionalfile1).Figures1,2,3and4depictthe mapsgeneratedfromthepredictivemodelfortheentire northernNamibiaregion,whileAdditionalfile1provides furtherdescriptionsanddatafromthemodelling.Table1 providespopulationweightedriskestimatesperhealth district.HumanmobilityAnalysesofradiusofgyration(Additionalfile2)show thatpopulationmovementsinNamibiafollowpatterns seenelsewhere[24,26,45]ofshorterdistancemovements beingsubstantiallymorecommonthanlargeronesand moreisolatedpopulationsgenerallytravellingfurther thanthoseindenselypopulatedareas(Additionalfile2). Acrossthe12-monthperiodexamined,abroadtrendof increasingphoneusageisevident(Figure6),withsome seasonalityinoverallmovementratesevident,including increasedactivityinDecember,justbeforethemainmalariatransmissionseason(Figure6).Spatially,movements followthemajortransportroutes,withthelargestamounts ofmovementseenwithintherelativelyhighlypopulated north-centralregion(Figure7).Table1presentssummariesofmobilitystatisticsbyhealthdistrict,withhealthdistrictsmappedinFigure5.Sources,sinksandcommunitiesofhumanandmalaria infectionmovementsSpatialheterogeneitiesinbothmovementpatternsand predictedmalariarisktranslateclearlyintovariationsin relativeratesofinfectionmovements,withphonecatchmentareasofstrongnetexportation(sources)located adjacenttoareasthatarenetsinks(Figure8).While therearedifferencesbetweenareasintermsofestimated netparasiteimportationandexportation,itisalsoclear thatmostofthenorthernregionconsistsofareasthat aresimultaneouslybothmajorsourcesandsinksof parasites(Figure9),ashighmovementratesdrive parasiteflowsacrossthere gion.Unsurprisingly,the north-centralborderregion,whichhassomeofthe highestpredictedrisksand largest,mostmobilepopulations,alsorepresentsthelargestsourceareaforthe country(Figure9).However,withpredictedmalaria riskconsistentacrossthisregion,heterogenitiesin movementpatternswithinitdrivevariationinriskconnectivity,meaningthattherearemanyregions,includingmostofthenorth,whicharebothnetimporters andhaveahighprobabilityofcasesseen(Figure9). ThesubstantialamountsoftravelfromWindhoekto themalariousnorthernregionsandback,andfromvisitorstoWindhoekfromthenorth,makethecapitalthe largestsinkarea(Figure9).Communitydetectionapplied totheweightednetworksofmovements,andmovements weightedbyrisk,betweenthe402phonecatchment areasresultedindifferingsetsofcommunitiesofstrongly connectedareasbeingfound(Additionalfile2),with spatialdifferencesalsoapparentbetweenreturningresidentsandvisitors(Additionalfile2).Table1summarizes communitymembershipbyhealthdistrict,withthose districtswithinthesamecommunitiesdisplayingstrongerlevelsofinternalconnectivitythroughmovements thanbetweendifferingcommunities,providingguidanceTatem etal.MalariaJournal 2014, 13 :52 Page7of15 http://www.malariajournal.com/content/13/1/52

PAGE 9

onwhichdistrictsshouldprioritizecoordinatingsurveillanceandcontroleffortsduetosubstantialpopulation andparasiteexchange.SpatialtargetingThequantificationofsource/sinkregionsshownin Figure3enablestheoreticalscenariosontheimpact Table1SummarystatisticsforeachhealthdistrictHealthdistrictPop 2011 %phone users Mean risk Mean RoG Meantrip length Meanno. trips Move comm Risk comm Popinrisk>50% andtop50source Meaneffect index Andara31,469250.0793.830.8672.1713897790.00382 Aranos27,669290.00107.370.9313.575800.00000 Eenhana104,313320.1560.400.6626.033100.00416 Engela127,931380.3445.230.3030.863100.00641 Gobabis95,225310.0078.361.113.5616500.00010 Grootfontein32,296580.0279.351.4312.8014600.00215 Karasburg19,946580.00163.642.552.635800.00000 KatimaMulilo80,460480.1678.500.717.084847730.00407 Keetmanshoop37,138600.00113.791.284.705500.00000 Khorixas17,839590.0084.482.354.3914400.00085 Luderitz24,890880.00202.861.9741.566100.00000 Mariental23,358600.00104.400.956.087400.00000 Nankudu38,601320.1171.040.5621.7748125050.00389 Nyangana23,109310.1157.500.2750.9113880640.00515 Okahandja70,058360.0067.630.494.867500.00000 Okahao35,674410.0348.860.3373.64144210.00165 Okakarara18,120570.0063.170.7020.5316500.00045 Okongo17,560460.12128.160.4663.762126280.00355 Omaruru32,738270.0070.620.8110.1615400.00000 Onandjokwe148,412360.0452.330.4318.258300.00192 Opuwo29,300400.0067.263.414.9812400.00155 Oshakati152,355760.0751.550.4017.869500.00339 Oshikuku120,363390.4251.400.2828.99107268290.00840 Otjiwarongo44,708680.0073.590.624.6614400.00012 Outapi63,890580.1659.000.6728.21117497030.00559 Outjo17,772590.0068.160.6310.1414400.00091 Rehoboth71,282300.0066.420.9410.157400.00000 Rundu67,743650.1383.980.4514.46138727770.01091 Swakopmund57,071860.00109.490.889.0315400.00000 Tsandi33,510200.2445.310.32121.8611761270.00364 Tsumeb20,5351000.0265.760.945.7314600.00582 Usakos12,174830.0077.600.6514.9115400.00000 Walvisbay57,3371070.00120.111.156.6615700.00000 Windhoek357,909900.0082.040.761.507500.00000HealthdistrictsaremappedinFigure 5 .Theshortenedcolumntitlesrefertothefollowing:Pop2011=numberofpeopleestimatedtoberesidingineach healthdistrictin2011;%Phoneusers=%ofPop2011populationthatwereestimatedtobecapturedintheCDRdataset,basedonnumbersofunique anonymoususerIDs;Meanrisk=meanpopulationweightedpredictedm alariacaseriskona0-1scale;MeanRoG=meanradiusofgyration[ 24 ]of movementsderivedfromthephonedata(SeeAdditionalfile 2 formoredetails);Meantriplength=mean lengthoftriptakenindaysawayfromhome phonecatchmentarea(SeeAdditionalfile 2 formoredetails);Meanno.trips=Meannumberoft ripstakenperyearawayfromhomephonecatchment area(SeeAdditionalfile 2 formoredetails);Movecomm=movementcommunitythatthemajor ityoftheareaofeachhealthdistrictbelongsto(Additional file 2 );Riskcomm=malariariskcommunitythatthemajorityofthe areaofeachhealthdistrictbelongsto(Additionalfile 2 );Popinrisk>50%andtop50 source=Numberofpeopleresidinginareaswhereriskvaluesare>0.5,andthatareinthetop50 ‘ source ’ regions(Figure 9 );Meaneffectindex=mean valueoftheeffectindexmappedinFigure 12 .Tatem etal.MalariaJournal 2014, 13 :52 Page8of15 http://www.malariajournal.com/content/13/1/52

PAGE 10

targetingofcontrolonmalariariskconnectivitytobe testedtoguideattackstrategies.Figures10and11demonstratehowdifferencesinmovementpatternscanmakea substantialdifferenceintermsofregionalimpactonrelativeratesofcaseimportationseenelsewherethrough interveningindifferentareas.InFigure10,ascenariois shownwherethepredictedmalariacaseriskatthephone catchmentareahighlightedisreducedtozero.Asthisis oneofthelargestsourceareas(Figure9),therelative impactofthisinterventionissubstantialacrossawide region,withmostimpactwithinthemalariamovement communityitbelongsto(Additionalfile2).Incontrast, thesameinterventioninaphonecatchmentareaofsimilarpopulationsizeandmalariarisk,butlowermobilityin termsofnumbersandrangeoftripsmadetoothercatchmentareas,showsasubstantiallysmallerimpact,both inmagnitudeandgeographicextentterms(Figure11). Theseinterventioneffectsonrelativeimpactsof Figure7 MovementtotalsbetweenhealthdistrictsoverOct2010-Sept2011period,withratesofmovementcolouredfromyellow (lowest)tored(highest). Figure8 Mapped ‘ sources ’ (netexporters)and ‘ sinks ’ (netimporters)ofmalariaimportationrisk. Areascolouredredareestimatedtobe netinfectionsourcesbasedonratesofmovementandmalariarisk,andthosecolouredbluearesinks,whilethosecolouredyellowareneither substantialnetexporternorimportersofinfections. Tatem etal.MalariaJournal 2014, 13 :52 Page9of15 http://www.malariajournal.com/content/13/1/52

PAGE 11

infectionexportationcanbesummarizedthroughasimple ‘ targeteffectiveness ’ metricthatmeasures,foreacharea, therelativelevelsofcaseimportationelsewherethatwould bestoppedifmalariariskwasreducedtozerointhearea. ThismetricismappedinFigure12andsummarizedby healthdistrictinTable1,andshowsaheterogenouspattern,indicatingthatthetargetingofsurveillanceand controlincertainareasmayhaveamuchlargerimpact onthesurroundingregionthaninotherneighbouring areas. Finally,Figure13demonstratestheutilityofthe combinedmappingandmovementquantificationapproachoutlinedhere,throughhighlightinghowhighrisk areasandpopulationscouldbeprioritizedforfurther investigation,surveillanceandcontrol.Existingnational guidelinescategorizetheentirenorthern ‘ zone1 ’ regionas thehigh-riskareawhereinterventionsshouldbefocused. Throughtherapidriskmappingapproach,areasandpopulationswithinitcanbehighlightedthatappeartobein particularlyhigherriskareasforcases.Thisrefinementreducesthepopulationtotargetfrom1.29millionresiding inthezone1region,to0.24millioninthepredictedhigher riskzones.Withinthesezones,populationmovements meanthatsomeareasarelikelytobelargerexporters (sources)ofinfections(Figures8and9)thanothers,and thetargetingofthesecanhaveabiggereffectonsurroundingareasthanthetargetingofsinks(Figures10, 11and12).Targetingonlythosepopulationsresiding Figure9 Mapped ‘ sources ’ (netexporters)and ‘ sinks ’ (netimporters)ofmalariaimportationrisk. Thelocationsofthetop50sourcesand sinksbyphonecatchmentareas. Figure10 Theinfluenceofconnectivitythroughhumanmobilityonthespatialimpactofinterventions. Thepercentagereductionin importationriskthroughreducingparasiteexportationnumberstozerointhephonecatchmentareamarkedinblue,whichisoneofthemajor sourceregionsinFigures8and9. Tatem etal.MalariaJournal 2014, 13 :52 Page10of15 http://www.malariajournal.com/content/13/1/52

PAGE 12

inthemajor ‘ source ’ areasofthehighriskzones,measuredinthiscasebythosephonecatchmentareasthat arethetop50largestsources(Figure9),furtherreducesthefocuspopulationto0.19million. Discussion Asmanycountriesreapthesuccessofrecentinvestmentsinmalariacontrolwithreportedcasesdeclining significantly,andre-orientatestrategiestowardselimination,parasitecarriagebyhumantravellersisrisingup nationalandglobalagendas[14,22,48,49].Inelimination settings,theimportationofmalariafromoutsideacountrybecomesthefocusofamalariacontrolprogramme, butintranationalhumanpopulationandmalariaparasite movementisanimportantpartofachievingelimination. Understandingthismovementshouldbeacriticalcomponentofthedesignofaneliminationstrategy,sinceit enablesprogrammestotargetresourcesinthemostefficientway,planattackstrategiesandensurethatcontextadaptedinterventionstrategiesareemployedacrossall high-riskareas.Pastdifficultiesinquantifyingandgainingabetterunderstandingofhumanmovementpatterns arebeingovercomethroughnewtechnologies[24,50] andherethepotentialofoneofthese,mobilephones,is outlinedinprovidingvaluableinformationthatcanbe integratedwithrapidcase-basedmalariariskmapping Figure11 Theinfluenceofconnectivitythroughhumanmobilityonthespatialimpactofinterventions. Thepercentagereductionin importationriskthroughreducingparasiteexportationnumberstozerointhephonecatchmentareamarkedinblue,whichisofsimilar populationsizeandriskleveltothefocusphonecatchmentareaofFigure10,buthaslowermovementrates. Figure12 Theinfluenceofconnectivitythroughhumanmobilityonthespatialimpactofinterventions. Mapofa ‘ targeteffectiveness ’ metric,whichmeasurestherelativereductioninimportationriskelsewherethroughcontrollingateachspecificlocation,withredlocations representingtheareaswherereducingparasiteexportationtozerohasthelargesteffectselsewhere,throughtogreen,whereminimaleffectsare seen.HealthdistrictnamesareshowninFigure5. Tatem etal.MalariaJournal 2014, 13 :52 Page11of15 http://www.malariajournal.com/content/13/1/52

PAGE 13

[16]toguidethedesignofdiseasecontrolandeliminationstrategies. Theanalysespresentedhereillustrateheterogeneities thatexistintermsofbothmalariariskandmobility acrossNamibia.Thecase-basedriskmappingresults (Figures1,2,3and4)revealtheconsistencyindriving factorsoftheprobabilityofcasesatthespatialscalesexaminedherebetweenthethreeregionsforwhichdata wereavailable(Additionalfile1),aswellastheaccuracy withwhichriskfactorsandareascanbedistinguished fromthelowerrisk ‘ background ’ conditions(Additional file1).Throughintegratingsuchhighresolutionrisk mappingwithCDRs,thetargetingofeliminationactivitiesthroughidentifyingaspectsofriskanalogousto boththe ‘ hotspots ’ and ‘ hotpops ’ concepts[49]couldbe undertakenifsystemflexibilityandcostsofundertaking thisallow,enablingthefocuseddeploymentoflimited resourcesinanattempttofocussurveillanceactivities andmaximizeimpact(Figure13).Inplanninganattack strategy,thinkingspatiallyandaccountingformobility couldbecritical – withamassdrugadministration (MDA)ormassscreenandtreat(MSAT)approach,reducingreceptivityinhightransmissionrisksinkscould beafocusthroughencouragingbednetuse,whilehigh transmissionsourcesareattacked(Figures1,2,3,4,8,9, 10,11and12).Suchanapproachwilllikelybemuch lesscostlyandoperationallydifficultthantryingto achieveblankethighcoverageofMDA/MSATinall high-riskareas(Figure13).Inpost-eliminationsettings, theframeworkpresentedhereprovidesguidancefortargetingsurveillancebyhighlightinghowareasthatare climatically,ecologicallyanddemographicallyreceptive totransmissionareconnectedbyhumanmovement (Figures8and9,Additionalfile2)andthroughexamininglikelysourcesandonwardmovementsfromlocal outbreaks.Itisclearthattheexportationofparasitesto otherlocationsisnotalwaysproblematicifthedestinationisnotreceptive,andtheapproachespresentedhere enabletheseparationofthese ‘ dead-end ’ movements frompossibleproblematicmovementstoreceptiveareas. Thedesignofstrategicplansforcontrolling,eliminating andpreventingmalariare-establishmentshould,therefore,ideallyaccountforhumanand,inturn,likelyparasitemovementpatterns,andtheanalysespresentedhere showthattoolsbuiltontheintegrationofdatasetsthat arereadilycollectedandstoredbycontrolprogrammes, satelliteoperatorsandmobilephonenetworkproviders canprovidethisvaluableinformationforprioritizing efforts. Whilsttheanalysespresentedoftheconnectivity betweenriskareasinamalariaeliminationsettinggo beyondpreviouswork,itisclearthatarangeofuncertaintiesremain.Manyofthosecrossingtheborderinto NamibiawillbecapturedbyphonedataduetoSIM cardswitching,butclearlyoneofthebiggestdrawbacksofsuchdataformobilityanalysesisthelackof Figure13 Malariariskzonemapsandthesizeofpopulationstotargetaccordingtothedifferentcategorizations. Therefinementofthe mappedareasshowshowthemethodcanbeusedtotargethigh-riskareasandpopulations,providingamethodforprioritizingthedeliveryof limitedresources. Tatem etal.MalariaJournal 2014, 13 :52 Page12of15 http://www.malariajournal.com/content/13/1/52

PAGE 14

cross-bordermovementratequantification.InfectionimportationsfromAngolaandotherneighbouringcountrieslikelyplayaroleintheepidemiologyofmalariain Namibia[48],andifthecommunitydetectionanalyses couldincludecross-bordermovementstheywouldlikely highlightthenorth-centralregionsasbeinginthesame communityassouth-centralAngolaandCaprivijoined withitssurroundingcountries,withmanyeconomicand familytiesacrosstheborderpromptingsignificantmovements[51]andcollaborationincontrolbeingvitalif eliminationistobeachieved[14,48].WhilephoneownershipandusageishighinNamibia,onlyacertainpercentageofthepopulationisbeingrepresentedbythe CDRsusedhere,andthesearepartiallybiasedtowards specificagegroupsandthericherandmoremobilesegmentsofthecountry[30,41](Additionalfile2).Moreover,thedemographicsanddailyactivitiesofnetwork subscribersremainrelativelyunknown(Additionalfile2), withdifferentgroupsandactivitieslikelypresentingsignificantlygreaterrisksofinfectionacquisitionthan others[22,47,52].However,recentanalysesonsimilardata inKenyasuggestthatthisisnotlikelytopresentasubstantialbiasinmobilityestimates[43]. Intermsoftheriskmappingundertaken,itremains clearthattheapproachidentifiesbroadareasofsuitabilityforfindingcasesbasedonecological,climatic, physicalanddemographicindicators,whichprovidesno guaranteeoffindingongoingtransmission.However,the cross-validationundertakensuggestsgoodperformance intermsofidentifyingareaswherecaseshaveoccurred (Additionalfile1),providingavaluabletoolforprioritizingareasforsurveillanceandfurtherinvestigation. Ideally,alternativemetricsoftransmission,suchasserologicalmarkers[53]shouldalsobeincorporatedasmore stablemeasuresoftransmissionandtoidentifyasymptomaticinfections,thus,betterquantifyingtruehotspots oftransmission,butsuchmeasuresarenotyetroutinely collected.Theutilizationoftrainingdatafromjustthree districtshere,wherealsospatialdifferencesintreatment seekingratesremainunknown,resultsinuncertaintiesin riskpredictionselsewhere,thoughtheaccuraciesinpredictionsandconsistencyinvariablesselectedastoppredictorsacrossthethreedistrictssuggeststhatthedrivers oftransmissionremainrelativelyconsistentcountrywide (Additionalfile1).Moreover,broadsimilaritiesoftheoutputstothemostrecentsurveillancedata[31]alsosuggests accuratemappingprospectively.Assessmentofthesensitivityofoutputspresentedheretovariabilityinqualityof surveillancesystemdatashouldrepresentanareaoffuture work,however.Ideally,informationonthereceptivity (thepropensitytoresultinonwardtransmissionfollowing animportedcase)ofeachareashouldformavaluableadditionalmetrictoimproveassessmentsoflocaltransmission risksfromcaseintroductions.Pre-controleraprevalence datahavebeenusedtodefinethisforthe1969-92period forNamibia[54,55],butsignificantdevelopment,populationgrowthandurbanizationoverrecentyears[42,56] havelikelychangedreceptivitysubstantially.Finally,the lackoftravelhistoriesinthecasedatausedraisesthepossibilitythatsomeinfectionswereacquiredawayfromtheir locationofresidence,thoughthestrongclusteringofcases isindicativeoflocaltransmissionandremovalofisolated casesleftoutputsunchanged(Additionalfile1). ThecontinuedupgradeoftheNamibiasurveillance system,aswellasthoseinothereliminationcountries, willbegintoprovidemorein-depthinformationoncases, enablingtheseparationoflikelylocal versus imported cases,aswellasthetravelhistoriesofimportedcases[57]. Theseimprovementsintype,qualityandquantityof surveillancedatawillinturnpresentopportunitiesfor theapplicationofimprovedspace-timestatisticalmappingapproachesandmathematicaltransmissionmodels toquantifyandaccountforuncertainties,aswellasthe estimationofpost-eliminationrisksofresurgence[23]. Asdatabecomemoreregularlyreported,acentralrepositoryintheformofanonlinemappingtoolislikelyto beanimportantassetforeliminationprograms[58,59]. Integratingintosuchatoolrapidcase-basedriskmapping thatcanbedynamicallyupdatedasnewdataarereported, toaccountforseasonalandinterannualvariations[16], wouldprovideusefulprioritizationforfurtherinvestigationsandsurveillanceactivities.Thelinkagetophonedata wouldthenprovidevaluableinformationonmobilityand connectivity.Further,combiningtheCDRswithother formsofmovementdata,suchascensus,surveyandsatellite[22,50,60],couldinformonthedemographics,drivers andseasonalityofmovements,aswellascross-border data,allofwhicharelackinginphonedata.Finally,many ofthemethodsoutlinedherearenotrestrictedtomalaria eliminationscenarios,withissuessuchasartemisinin resistancespread[61,62],vaccine-preventablechildhood illnesses[63],andtheeliminationofotherdiseases[64] alsoreliantonanunderstandingofmovementdynamics.AdditionalfilesAdditionalfile1: Case-basedmalariariskmapping – additional details. Additionalinformationonthedatasets,methodsandresultsfor thecase-basedriskmapping. Additionalfile2: Mobilephonecalldetailrecords – additional details. Furtherinformationonphoneownership,mobilephonenetwork geographyandmobilitypatternsinNamibia. Competinginterests Theauthorsdeclarethattheyhavenocompetinginterests. Authors ’ contributions AJT,CL,JC,ZH,DKP,JCandDKdesignedthestudy.CN,PU,CLandAJT undertookmalariadatacollectionandprocessing.GB,ZH,UKandDKP processedthemobilephoneandhouseholdsurveydata.AJT,ZHandUKTatem etal.MalariaJournal 2014, 13 :52 Page13of15 http://www.malariajournal.com/content/13/1/52

PAGE 15

undertookthemalariariskmappingandphonedataintegration.Allauthors contributedtothewritingofthemanuscriptandhavereadandapproved thefinalversion. Acknowledgements TheworkwasundertakenunderadatasharingagreementbetweenMTC Namibia,theNamibianNationalVector-borneDiseasesControlProgramme andtheClintonHealthAccessInitiative.TheauthorsaregratefultoMTCfor sharingtheirdataandhelpwithextractions.Thisworkrepresentspartofthe HumanMobilityMappingProject(http://www.thummp.org),Flowminder (www.flowminder.org)andtheWorldPoppopulationmappingproject (http://www.worldpop.org.uk).AJT&DLSacknowledgefundingsupportfrom theRAPIDDprogramoftheScienceandTechnologyDirectorate,DepartmentofHomelandSecurity,andtheFogartyInternationalCenter,National InstitutesofHealth,andarealsosupportedbygrantsfromNIH/NIAID (U19AI089674)andtheBillandMelindaGatesFoundation(#49446and #1032350).Thefundershadnoroleinstudydesign,datacollectionand analysis,decisiontopublish,orpreparationofthemanuscript. Authordetails1DepartmentofGeographyandEnvironment,UniversityofSouthampton, Southampton,UK.2FogartyInternationalCenter,NationalInstitutesofHealth, Bethesda,MD,USA.3DepartmentofGeography,UniversityofFlorida, Gainesville,FL,USA.4EmergingPathogensInstitute,UniversityofFlorida, Gainesville,FL,USA.5NationalVector-borneDiseaseControlProgramme, Windhoek,Namibia.6DepartmentofComputerScience,UniversityofFlorida, Gainesville,FL,USA.7ClintonHealthAccessInitiative,Boston,MA,USA.8DepartmentofEpidemiology,JohnsHopkinsBloombergSchoolofPublic Health,Baltimore,MD,USA.9MobileTelecommunicationsLimited,Windhoek, Namibia. Received:22November2013Accepted:3February2014 Published:10February2014 References1.BaratLM: Fourmalariasuccessstories:howmalariaburdenwas successfullyreducedinBrazil,Eritrea,IndiaandVietnam. AmJTropMed Hyg 2006, 74: 12 – 16. 2.BarnesKI,DurrheimDN,LittleF,JacksonA,MehtaU,AllenE,DlaminiSS, TsokaJ,BredenkampB,MthembuDJ,WhiteNJ,SharpBL: Effectof artemether-lumefantrinepolicyandimprovedvectorcontrolonmalaria burdeninKwaZulu-Natal,SouthAfrica. PLoSMed 2005, 2: e330. 3.BhattaraiA,AliAS,KachurSP,MartenssonA,AbbasAK,KhatibR,Al-MafazyAW, RamsanM,RotllantG,GerstenmaierJF,MolteniF,AbdullaS,MontgomerySM, KanekoA,BjorkmanA: Impactofartemisinin-basedcombinationtherapyand insecticide-treatednetsonmalariaburdeninZanzibar. PLoSMed 2007, 4: e309. 4.CeesaySJ,Casals-PascualC,Ers kineJ,AnyaSE,DuahNO,FulfordAJ, SesaySS,AbubakarI,DunyoS,SeyO,PalmerA,FofanaM,CorrahT, BojangKA,WhittleHC,Gr eenwoodBM,ConwayDJ: Changesinmalariaindices between1999and2007inTheGamb ia:aretrospectiveanalysis. Lancet 2008, 372: 1545 – 1554. 5.FeganGW,NoorAM,AkhwaleWS,CousensS,SnowRW: Effectof expandedinsecticide-treatedbednetcoverageonchildsurvivalinrural Kenya:alongitudinalstudy. Lancet 2007, 370: 1035 – 1039. 6.KleinschmidtI,SchwabeC,BenaventeLE,TorrezM,RidlFC,SeguraJL, EhmerP,NchamaGN: Markedincreaseinchildsurvivalafterfouryearsof intensivemalariacontrol. AmJTropMedHyg 2009, 80: 882 – 888. 7.NyarangoPM,GebremeskelT,MebrahtuG,MufundaJ,AbdulmuminiU, OgbamariamA,KosiaA,GebremichaelA,GunawardenaD,GhebratY, OkbaldetY: Asteepdeclineofmalariamorbidityandmortalitytrendsin Eritreabetween2000and2004:theeffectofcombinationofcontrol methods. MalarJ 2006, 5: 33. 8.OkiroEA,HaySI,GikandiPW,Sharif SK,NoorAM,PeshuN,MarshK,SnowRW: ThedeclineinpaediatricmalariaadmissionsonthecoastofKenya. MalarJ 2007, 6: 151. 9.O'MearaWP,BejonP,MwangiTW,OkiroEA,PeshuN,SnowRW,NewtonCR, MarshK: Effectofafallinmalariatransmissiononmorbidityandmortalityin Kilifi,Kenya. Lancet 2008, 372: 1555 – 1562. 10.TeklehaimanotHD,TeklehaimanotA,KiszewskiA,RampaoHS,SachsJD: MalariainSoTomandPrincipe:Onthebrinkofeliminationafter threeyearsofeffectiveantimalarialmeasures. AmJTropMedHyg 2009, 80: 133 – 140. 11.RollBackMalaria: GlobalMalariaActionPlan. Geneva:RollBackMalaria partnership;2008. 12.FeachemRGA,PhillipsAA,TargettGA(Eds):ShrinkingtheMalariaMap:A ProspectusonMalariaElimination. SanFrancisco:TheGlobalHealthGroup, GlobalHealthSciences,UniversityofCalifornia;2009. 13.WorldHealthOrganization: WorldMalariaReport. Geneva,Switzerland: WorldHealthOrganization;2011. 14.CotterC,SturrockHJ,HsiangMS,LiuJ,PhillipsAA,HwangJ,GueyeCS, FullmanN,GoslingRD,FeachemRG: Thechangingepidemiologyofmalaria elimination:newstrategiesfornewchallenges. Lancet 2013, 382: 900 – 911. 15.MoonenB,CohenJM,SnowRW,SlutskerL,DrakeleyC,SmithDL, AbeyasingheRR,RodriguezMH,MaharajR,TannerM,TargettG: Operationalstrategiestoachieveandmaintainmalariaelimination. Lancet 2010, 376: 1592 – 1603. 16.CohenJM,DlaminiSS,NovotnyJM,KandulaD,KuneneS,TatemAJ: Rapid case-basedmappingofseasonalmalariatransmissionriskforstrategic eliminationplanninginSwaziland. MalarJ 2013, 12: 61. 17.GethingPW,ElyazarIR,MoyesCL,SmithDL,BattleKE,GuerraCA,PatilAP, TatemAJ,HowesRE,MyersMF,GeorgeDB,HorbyP,WertheimHF,PriceRN, MuellerI,BairdJK,HaySI: Alongneglectedworldmalariamap: Plasmodium vivax endemicityin2010. PLoSNeglTropDis 2012, 6: e1814. 18.GethingPW,PatilAP,SmithDL,GuerraCA,ElyazarIR,JohnstonGL,TatemAJ, HaySI: Anewworldmalariamap: Plasmodiumfalciparum endemicityin 2010. MalarJ 2011, 10: 378. 19.HaySI,SmithDL,SnowRW: Measuringmalariaendemicityfromintense tointerruptedtransmission. LancetInfectDis 2008, 8: 369 – 378. 20.KuneneS,PhillipsAA,GoslingRD,KandulaD,NovotnyJM: Anational policyformalariaeliminationinSwaziland:afirstforsub-SaharanAfrica. MalarJ 2011, 10: 313. 21.USAID: EvaluationReport:MalariaEradicationProgram,Haiti:January17-27, 1972. PortauPrince,Haiti:USAID;1972. 22.PindoliaDK,GarciaAJ,WesolowskiA,SmithDL,BuckeeCO,NoorAM,SnowRW, TatemAJ: Humanmovementdataformalariacontrolandelimination strategicplanning. MalarJ 2012, 11: 205. 23.CohenJM,SmithDL,CotterC,WardA,YameyG,SabotOJ,MoonenB: Malariaresurgence:asystematicreviewandassessmentofitscauses. MalarJ 2012, 11: 122. 24.GonzalezMC,HidalgoCA,BarabasiAL: Understandingindividualhuman mobilitypatterns. Nature 2008, 453: 779 – 782. 25.TatemAJ,QiuY,SmithDL,SabotO,AliAS,MoonenB: Theuseofmobile phonedatafortheestimationofthetravelpatternsandimportedPlasmodiumfalciparum ratesamongZanzibarresidents. MalarJ 2009, 8: 287. 26.SongC,QuZ,BlummN,BarabasiAL: Limitsofpredictabilityinhuman mobility. Science 2010, 327: 1018 – 1021. 27.BengtssonL,LuX,ThorsonA,GarfieldR,vonSchreebJ: Improvedresponseto disastersandoutbreaksbytrackingpopulationmovementswithmobile phonenetworkdata:apost-earthquakegeospatialstudyinHaiti. PLoSMed 2011, 8: e1001083. 28.MoonenB,CohenJM,TatemAJ,CohenJ,HaySI,SabotO,SmithDL: A frameworkforassessingthefeasibilityofmalariaelimination. MalarJ 2010, 9: 322. 29.LeMenachA,TatemAJ,CohenJM,HaySI,RandellH,PatilAP,SmithDL: Travelrisk,malariaimportationandmalariatransmissioninZanzibar. Sci Rep 2011, 1: 93. 30.WesolowskiA,EagleN,TatemAJ,SmithDL,NoorAM,SnowRW,Buckee CO: Quantifyingtheimpactofhumanmobilityonmalaria. Science 2012, 338: 267 – 270. 31.NamibiaNationalVector-b orneDiseaseProgramme: MonthlyHealth ManagementInformationSystemReports. Windhoek:NamibiaNational Vector-borneDiseaseProgramme;2012. 32.NamibiaNationalVector-borne DiseasesControlProgramme: National MalariaPolicy. TheRepublicofNamibia:MinistryofHealthandSocial Services,DirectorateofSpecialprogrammes;2013. 33.LinardC,GilbertM,SnowRW,NoorAM,TatemAJ: Populationdistribution, settlementpatternsandaccessibilityacrossAfricain2010. PLoSONE 2012, 7: e31743. 34.ElithJ,GrahamCH,AndersonRP,DudikM,FerrierS,GuisanA,HijmansRJ, HuettmannF,LeathwickJR,LehmannA,LiJ,LohmannLG,LoiselleBA,Tatem etal.MalariaJournal 2014, 13 :52 Page14of15 http://www.malariajournal.com/content/13/1/52

PAGE 16

ManionG,MoritzC,NakamuraM,NakazawaY,OvertonJM,PetersonAT, PhillipsSJ,RichardsonK,Scachetti-Per eiraR,SchapireRE,SoberonJ,WilliamsS, WiszMS,ZimmermannNE: Novelmethodsimprovepredictionofspecies' distributionsfromoccurrencedata. Ecography 2006, 29: 129 – 151. 35.PhillipsSJ,DudikM: ModelingofspeciesdistributionswithMaxent:new extensionsandacomprehensiveevaluation. Ecography 2008, 31: 161 – 175. 36.BreimanL: Randomforests. MachLearn 2001, 45: 5 – 32. 37.RDevelopmentCoreTeam: R:ALanguageandEnvironmentforStatistical Computing. ;2009. 38.De'athG,FabriciusKE: Classificationandregressiontrees:apowerfulyet simpletechniqueforecologicaldataanalysis. Ecology 2000, 81: 3178 – 3192. 39.PearceF,FerrierS: Evaluatingthepredictiveperformanceofhabitat modelsdevelopedusinglogisticregression. EcolModel 2000, 133: 225 – 245. 40.RogersDJ: Modelsforvectorsandvector-bornediseases. AdvParasitol 2006, 62: 1 – 35. 41.CommunicationsRegulatoryAuthorityofNamibia: UniversalServiceBaseline Study. Windhoek:CommunicationsRegulatoryAuthorityofNamibia;2011. 42.UnitedNationsPopulationDivision: WorldPopulationProspects,2012 Revision. NewYork:UnitedNations;2012. 43.WesolowskiA,EagleN,NoorAM,SnowRW,BuckeeCO: Theimpactof biasesinmobilephoneownershiponestimatesofhumanmobility. JR SocInterface 2013, 10: 20120986. 44.BalkDL,DeichmannU,YetmanG,PozziF,HaySI,NelsonA: Determining globalpopulationdistribution:methods,applicationsanddata. Adv Parasitol 2006, 62: 119 – 156. 45.SiminiF,GonzalezMC,MaritanA,BarabasiAL: Auniversalmodelfor mobilityandmigrationpatterns. Nature 2012, 484: 96 – 100. 46.BlondelVD,GuillaumeJL,LambiotteR,LefebvreE: Fastunfoldingof communitiesinlargenetworks. JStatMechTheorExp 2008, 2008: P10008.47.PindoliaDK,GarciaAJ,HuangZ,SmithDL,AleganaVA,NoorAM,SnowRW, TatemAJ: Thedemographicsofhumanandmalariamovementand migrationpatternsinEastAfrica. MalarJ 2013, 12: 397. 48.TatemAJ,SmithDL: Internationalpopulationmovementsandregional Plasmodiumfalciparum malariaeliminationstrategies. ProcNatlAcadSci USA 2010, 107: 12222 – 12227. 49.SturrockHJ,HsiangMS,CohenJM,SmithDL,GreenhouseB,BousemaT, GoslingRD: Targetingasymptomaticmalariainfections:active surveillanceincontrolandelimination. PLoSMed 2013, 10: e1001467. 50.StoddardST,MorrisonAC,Vazquez-ProkopecGM,SoldanVP,KochelTJ, KitronU,ElderJP,ScottTW: Theroleofhumanmovementinthetransmission ofvector-bornepathogens. PLoSNeglTropDis 2009, 3: e481. 51.NangulahSMW,NickanorNM: NorthernGateway:Cross-BorderMigration BetweenNamibiaandAngola,Souther nAfricaMigrationProjectReport. SouthAfrica:CapeTown;2005. 52.TatemAJ,AdamoS,BhartiN,BurgertCR,CastroM,DorelienA,FinkG, LinardC,MendelsohnJ,MontanaL,MontgomeryMR,NelsonA,NoorAM, PindoliaD,YetmanG,BalkD: Mappingpopulationsatrisk:improving spatialdemographicdataforinfectiousdiseasemodelingandmetric derivation. PopulHealthMetr 2012, 10: 8. 53.BousemaT,DrakeleyC,GesaseS,HashimR,MagesaS,MoshaF,OtienoS, CarneiroI,CoxJ,MsuyaE,KleinschmidtI,MaxwellC,GreenwoodB,RileyE, SauerweinR,ChandramohanD,GoslingR: Identificationofhotspotsof malariatransmissionfortargetedmalariacontrol. JInfectDis 2010, 201: 1764 – 1774. 54.NoorAM,AleganaVA,KamwiRN,HansfordCF,NtomwaB,KatokeleS, SnowRW: Malariacontrolandtheintensityof Plasmodium falciparum transmissioninNamibia1969-1992. PLoSOne 2013, 8: e63350. 55.NoorAM,UusikuP,KamwiRN,KatokeleS,NtomwaB,AleganaVA,Snow RW: Thereceptiveversuscurrentrisksof Plasmodiumfalciparum transmissioninNorthernNamibia:implicationsforelimination. BMC InfectDis 2013, 13: 184. 56.UnitedNationsPopulationDivision: WorldUrbanizationProspects,2011 Revision. NewYork:UnitedNations;2011. 57.SturrockHJ,NovotnyJM,KuneneS,DlaminiS,ZuluZ,CohenJM,Hsiang MS,GreenhouseB,GoslingRD: Reactivecasedetectionformalaria elimination:real-lifeexperiencefromanongoingprograminSwaziland. PLoSOne 2013, 8: e63830. 58.KellyGC,HaleE,DonaldW,BatariiW,B ugoroH,NausienJ,SmaleJ,PalmerK, BobogareA,TaleoG,VallelyA,TannerM,VestergaardLS,ClementsAC: A high-resolutiongeospatialsurveill ance-responsesystemformalaria eliminationinSolomon IslandsandVanuatu. MalarJ 2013,12: 108. 59.ClementsAC,ReidHL,KellyGC,HaySI: Furthershrinkingthemalariamap: howcangeospatialsciencehelptoachievemalariaelimination? Lancet InfectDis 2013, 13: 709 – 718. 60.BhartiN,TatemAJ,FerrariMJ,GraisRF,DjiboA,GrenfellBT: Explaining seasonalfluctuationsofmeaslesinNigerusingnighttimelightsimagery. Science 2011, 334: 1424 – 1427. 61.FleggJA,PatilAP,VenkatesanM,RoperC,NaidooI,HaySI,SibleyCH, GuerinPJ: Spatiotemporalmathematicalmodellingofmutationsofthe dhps geneinAfrican Plasmodiumfalciparum MalarJ 2013, 12: 249. 62.PearceRJ,PotaH,EveheMS,BaelH,Mombo-NgomaG,MalisaAL,OrdR, InojosaW,MatondoA,DialloDA,MbachamW,vandenBroekIV,SwarthoutTD, GetachewA,DejeneS,GrobuschMP,NjieF,DunyoS,KwekuM,Owusu-AgyeiS, ChandramohanD,BonnetM,GuthmannJP,ClarkeS,BarnesKI,StreatE, KatokeleST,UusikuP,AgboghoromaCO,ElegbaOY, etal : Multiple originsandregionaldispersalofresistantdhpsinAfrican Plasmodium falciparum malaria. PLoSMed 2009, 6: e1000055. 63.FerrariMJ,GraisRF,BhartiN,ConlanA J,BjornstadON,WolfsonLJ,GuerinPJ, DjiboA,GrenfellBT: Thedynamicsofmeaslesinsub-SaharanAfrica. Nature 2008, 451: 679 – 684. 64.KlepacP,MetcalfJE,McLeanAR,HampsonK: Towardstheendgameand beyond:complexitiesandchallengesfortheeliminationofinfectious diseases. PhilTransRSocLondBBiolSci 2013, 368: 20120137.doi:10.1186/1475-2875-13-52 Citethisarticleas: Tatem etal. : Integratingrapidriskmappingand mobilephonecallrecorddataforstrategicmalariaelimination planning. MalariaJournal 2014 13 :52. Submit your next manuscript to BioMed Central and take full advantage of: € Convenient online submission € Thorough peer review € No space constraints or color “gure charges € Immediate publication on acceptance € Inclusion in PubMed, CAS, Scopus and Google Scholar € Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit Tatem etal.MalariaJournal 2014, 13 :52 Page15of15 http://www.malariajournal.com/content/13/1/52