改进新型流行病病原体检测的策略-有前景的病原体不诊断检测工作流程的成本与检测性能

ResearchReport

ALEXDEMARSH,PEDRONASCIMENTODELIMA,CASSIDYNELSON,JAVIERROJASAGUILERA,NATHANDUARTE,SELLANEVO,HENRYH.WILLIS

Strategiesto

ImproveDetection

ofNovelPandemic

Pathogens

CostVersusDetectionPerformanceforPromising

Pathogen-AgnosticDetectionWorkflows

RR-A3704-1

ThispublicationhascompletedRAND’sresearchquality-assurance

processbutwasnotprofessionallycopyedited.

Formoreinformationonthispublication,visit

/t/RRA3704-1.

AboutRAND

RANDisaresearchorganizationthatdevelopssolutionstopublicpolicychallengestohelpmakecommunitiesthroughouttheworldsaferandmoresecure,healthierandmoreprosperous.RANDisnonprofit,nonpartisan,andcommittedtothepublicinterest.TolearnmoreaboutRAND,visit

.

ResearchIntegrity

Ourmissiontohelpimprovepolicyanddecisionmakingthroughresearchandanalysisisenabledthroughourcorevaluesofqualityandobjectivityandourunwaveringcommitmenttothehighestlevelofintegrityandethicalbehavior.Tohelpensureourresearchandanalysisarerigorous,objective,andnonpartisan,wesubjectourresearchpublicationstoarobustandexactingquality-assuranceprocess;avoidboththeappearanceandrealityoffinancialandotherconflictsofinterestthroughstafftraining,projectscreening,andapolicyofmandatorydisclosure;andpursuetransparencyinourresearchengagementsthroughourcommitmenttotheopenpublicationofourresearchfindingsandrecommendations,disclosureofthesourceoffundingofpublishedresearch,andpoliciestoensureintellectualindependence.Formoreinformation,visit

/about/research-integrity.

RAND’spublicationsdonotnecessarilyreflecttheopinionsofitsresearchclientsandsponsors.

PublishedbytheRANDCorporation,SantaMonica,Calif.

©2025RANDCorporation

RANDoisaregisteredtrademark.

LimitedPrintandElectronicDistributionRights

Thispublicationandtrademark(s)containedhereinareprotectedbylaw.ThisrepresentationofRANDintellectualpropertyisprovidedfornoncommercialuseonly.Unauthorizedpostingofthispublicationonlineisprohibited;linkingdirectlytoitswebpageonisencouraged.PermissionisrequiredfromRANDtoreproduce,orreuseinanotherform,anyofitsresearchproductsforcommercialpurposes.Forinformationonreprintandreusepermissions,visit

/about/publishing/permissions

.

AboutThisResearchReport

iii

Thepossibilityofapandemiccausedbyanovelpathogenmotivatesthedevelopmentand

deploymentofbiosurveillancesystemscapableofrapidlyrecognizingnovelthreats.Toinformdecisionsthatenablethenextgenerationofbiosurveillancesystemstorapidlydetectoutbreaksofnovelpathogens,ourteamevaluatedthecostanddetectionperformanceofthreepromising

pathogen-agnosticdetectionstrategies.Thesestrategiesemployedcombinationsofnovel

detectiontechnologiesinanidealizedsettingofdeploymentatmilitarybasesupportedbya

dedicatedmedicalfacility.Thisreportpresentsthetechnologiesandsystemsevaluatedand

detailstheresultsofasimulationstudydesignedtoestimatethetotalsystemcostversusits

detectionperformance,asmeasuredbytheproportionofpopulationinfectedatthetimeofthe

detectionandthetimefromemergencetodetection.Givenourfocusonearlydetection,this

analysisdoesnotprovidecomprehensivecosteffectivenessofthesesystemsasouranalysisdoesnotincludeactions,costs,andoutcomesfollowingdetectionofadiseaseoutbreak,suchascostsofdiseasedeathsandillnessorthecostsandeffectivenessofpandemicresponse.Theresultswillbeofinteresttodecisionmakersconsideringdeployingsuchasystemandthemethodscan

informanalysisofothertypesofbiosurveillancesystemstoserveotherpopulationsandcommunities.

CenteronAI,Security,andTechnology

RANDGlobalandEmergingRisksisadivisionofRANDthatdeliversrigorousand

objectivepublicpolicyresearchonthemostconsequentialchallengestocivilizationandglobalsecurity.Thisworkwasundertakenbythedivision’sCenteronAI,Security,andTechnology,

whichaimstoexaminetheopportunitiesandrisksofrapidtechnologicalchange,focusingon

artificialintelligence,security,andbiotechnology.Formoreinformation,contact

cast@

.

Funding

ThisresearchwasindependentlyinitiatedandconductedwithintheCenteronAI,Security,andTechnologyusingincomefromoperationsandgiftsandgrantsfromphilanthropic

supporters.Acompletelistofdonorsandfundersisavailableat

/CAST

.RANDclients,donors,andgrantorshavenoinfluenceoverresearchfindingsorrecommendations.

Acknowledgments

TheauthorsthankToddHelmus(RAND)forextensiveguidanceonanearlyversionofthisreport,andLisaColabella(RAND)forhersignificantcontributiontocostanalysis.Wefurther

iv

thankForrestCrawford(RAND)andJacobSwett(BlueprintBiosecurity)forhelpfulfeedback

throughout,andCaseyAveggio(RAND)forsupport.WealsothankCharlesChiu(UniversityofCalifornia,SanFrancisco)andRaffaeleVardavas(RAND)fortheircarefulandconstructive

review.

Summary

v

Detectingnovelpathogensisasignificantchallengeforbiosurveillancesystems,inpart

becausetheyrelyonpathogen-specificdetectiontechnologies.Decisionmakersplanningto

improvebiosurveillancesystemsusingpathogenagnostictechnologieslackevidenceontheirexpectedcostsandtheexpecteddetectionperformanceofalternativesurveillancesystem

designs.Investmentinimprovedbiosurveillancesystemsislimitedbyalackofclarityoncostversusperformance,aswellascompleximplementationchoices.

Approach

Thisreportintroducesamodelofthreebiosurveillancestrategies,referredtoasSyndromic,Wearable,andEnvironmental.Eachoftheseisapathogen-agnosticworkflowdesignedtodetectnovelthreatsviametagenomicnext-generationsequencing(mNGS).TheSyndromicand

Wearablestrategiesareinitiatedbyasignalfromsymptomaticorphysiologicalsensors

(respectively),whiletheEnvironmentalstrategyrunscontinuouslyandisintensified

responsivelybasedonexternalintelligencefromevent-basedsurveillance.Allstrategiesaredescribedfullyinthemaintext.First,weparameterizethismodeltoemulatethedetectionofwildtypeSARS-Cov-2.Then,weexplorehowchangesinpathogencharacteristicsand

technologyperformancechangekeydetectionandcostmeasures.Weinformourmodelwithparametersderivedfromtheliterature,andpubliclyavailableinformationonfixed,operating,andvariablecostsofdiseasesurveillanceactivities.

Thisreportprovidesinsightintothetradeoffsinherenttothedesignofbiosurveillance

systemsbysimulatingthreealternativestrategies.Ourmodelingapproachrequiresasetof

assumptionsaboutthepathogenbeingdetected(i.e.,transmissibilityanddiseaseprogression),

thedetectiontechnology(i.e.,sensitivityofalternativedetectionsystems)andhowitwouldbe

usedinareal-worldsetting.Theseassumptionsaredocumentedinourmainreportandits

technicalappendix.Thissimplifiedsettingandtreatmentofstrategiesasdiscretechoicesis

intendedtocapturekeyfeaturesofadeployedmilitarypopulationandprovideroughcostversusperformanceanalysisasoneinputforchoosinganovelsurveillancesystem.Acrossallscenarios,weassumetheexistenceofabaselinebiomedicalcapability,includingaccesstostandard

diagnostics,andbasicoutbreakinvestigationprocedures.Areal-worldimplementationwouldbuildonthisexistinginfrastructure,andlikelydeploythestrategiesconsideredherein

combinationratherthanatomically.

vi

KeyFindings

•TheEnvironmentalstrategyaffordedthelargestimprovementsindetectionperformanceinourbase-casescenario,followedbythewearablestrategy.

•Thesuperiorityoftheenvironmentalandwearablestrategiesoverthesyndromicstrategydegradesinscenarioswherethediseasehashightransmissibility,theproportionof

individualswhoareasymptomaticislowandwhenthediseasehasafastlatentperiod.

Recommendations

•Alimiteddeploymentwouldallowoperatingcoststobeestimatedwithmoreprecision,provideanopportunitytocollectlogisticalandoperationaldata,estimatedetection

performanceunderreal-worldconditions,andexploredatafusionwiththelargerbiosurveillanceenterprise.

•Thereissignificantuncertaintyinthereal-worldperformanceofnext-generation

sequencingtechnologies,whichcouldbegreatlyreducedviaalimiteddeploymentofoneofthestrategiesexploredinthisreport,oracombinationthereof.

Contents

vii

AboutThisResearchReport iii

Summary v

FiguresandTables viii

StrategiestoImproveDetectionofNovelPandemicPathogens:CostVersusDetection

PerformanceforPromisingPathogen-AgnosticDetectionWorkflows 1

Background 1

ThreeTechnologiesAssessedinThisStudy 2

Methods 3

Results 6

DetectionPerformanceandCostsVersusPathogenLatencyPeriods 8

Discussion 10

AppendixA.AssessedStrategies 12

AppendixB.SimulationModel 14

Abbreviations 28

References 29

AbouttheAuthors 33

FiguresandTables

viii

Figures

Figure1.DetectionPerformanceandCostwithVaryingReproductionNumbers 8

Figure2.DetectionPerformanceandCostwithVaryingLatentPeriod 9

FigureA.1.ConceptualIllustrationofDetectionstrategies 13

FigureB.1.CompartmentModelStructure 14

Tables

Table1.PerformanceofSurveillanceStrategiesinBase-CaseScenario 7

TableB.1.TransmissionModelParameters 16

TableB.2.SyndromicStrategyParameters 17

TableB.3.Wearablesstrategyparameters 19

TableB.4.EnvironmentalStrategyParameters 21

TableB.5.CostEstimatesfortheSyndromicStrategyperFacility 23

TableB.6.CostEstimatesfortheWearablesStrategyperFacility 24

TableB.7.CostEstimatesfortheEnvironmentalStrategyperFacility 25

TableB.8.SyndromicSurveillancePerformanceinScenarioDiscoveryAnalysis 25

TableB.9.PRIMBoxesforSyndromicDetectionWithin15DaysofWearable 26

TableB.10.PRIMBoxesforSyndromicDetectionWithin15DaysofEnvironmental 26

TableB.11.CARTClassificationPerformanceforRelativeDetection 26

TableB.12.ParameterImportanceRankingforDetectionPerformance(RandomForest) 27

1

StrategiestoImproveDetectionofNovelPandemicPathogens:CostVersusDetectionPerformanceforPromisingPathogen-

AgnosticDetectionWorkflows

Background

Thepossibilityofapandemiccausedbyanovelpathogenmotivatesthedevelopmentanddeploymentofbiosurveillancesystemscapableofrapidlyrecognizingnovelthreats.Pandemicscanbeinitiatedbyanaturallyemergingzoonosis,alaboratoryorresearchaccident,or

deliberatelyreleased.Recognizingthesignificanceoftheearliestcasesinanascentepidemicisdifficultforexistingbiosurveillancesystems,evenforfamiliarpathogens.Thechallengeis

significantlygreaterfornovelpathogensbecausepathogen-specificdetectiontechnologies

requireasignificanttimeinvestmentbeforetheycanbeadaptedtoanewthreat(Mortonetal.,2024).Thislaghashistoricallybeenovercomeviasyndromicsurveillance,whichentails

analyzingclustersofsymptomreportswithoutrequiringspecificdiagnosis(Henning,2004).

Still,anewoutbreakmaynotbesuspecteduntilmanyseverelyillcasespresenttoahealthservice,bywhichstageseveralhundredorthousandsmorecouldhavebeeninfected.

However,evenwhenrobustmonitoringandsyndromicsurveillancesystemsareinplace,

detectingemergingpathogenoutbreakshashistoricallyentailedadditionaldelays.Inaglobal

analysisofoutbreakreportsinvolvingbothknownandunknownemergingpathogens,the

mediantimebetweenoutbreakstartanddiscoverywas20days,withlaboratoryconfirmation

takingamedianof36days(Klubergetal.,2016).Deliberatelyreleasedengineeredagents

intendedtoevadecurrentsurveillancecouldpotentiallyhaveevenlongertime-to-detectionusingcurrentsystems.

Newclinicalandmoleculartestingtechnologiescanbeintegratedintobiosurveillance

systemstoprovideearlywarningofnascentepidemics,andifcoupledwithrobustresponse

measurescouldreducetheprobabilityofdevelopingintopandemics.Newdetectiontechnologiesareatvariousstagesoftechnologyreadiness.Somearecommerciallyavailableandwidelyused,suchasmultiplexPCRandsimilarbroad-scopemoleculartestsforclinicalsamples.Consumerinnovationsarecreatingnewusesandapplicationssuchaswearablebiometricsand

physiologicalsensors.Otheremergingtechnologiesareinearly-stageadoptionatspecializedlaboratories,suchaspathogen-agnosticmetagenomicanalysisofclinicalandenvironmental

samples.Researchsuggestspromiseinincorporatingpathogen-agnostictoolsinto

biosurveillance(Bassietal.,2022;Bohletal.,2022;Dengetal.,2020;Gardy&Loman,2018;Gauthieretal.,2023;Govenderetal.,2021;Milleretal.,2013;Mohsinetal.,

2021).Implementingthesetechnologiesinreal-worldbiosurveillancesystemspresentscost-

2

effectivenesstradeoffs,yettheresearchonthecost-effectivenessofthosesystemsisscant.cost-effectivenesstradeoffs,yettheresearchonthecost-effectivenessofthosesystemsisscant.

Toinformdecisionsthatenablethenextgenerationofbiosurveillancesystemstorapidly

detectoutbreaksofnovelpathogens,ourteamevaluatedthecostanddetectionperformanceofthreepromisingpathogen-agnosticdetectionstrategiesemployingnoveldetectiontechnologies.Thespecificidealizedsettingforthisanalysisisdeploymentatmilitarybasesupportedbya

dedicatedmedicalfacility.Thissitewasselectedbecausetwocharacteristicsmakeita

compellingsiteforapilotdeploymentofsuchasystem.First,thesiteinvolvesawell

characterized“closed”populationthatisreasonablyself-contained,whichsignificantly

simplifiesanalysiswhilestillprovidingreal-worldrelevance.Second,inthiscontext,the

militaryhasbroadauthoritiestoimplementnovelbiosurveillancesystems,andatrackrecordofsignificantinvestmentinnoveltechnologiestoprotectthewarfighter.Narrativesofthethree

strategiesareprovidedinAppendixAanddescribedinmoredetailinthefollowingsectionsofthisreport.

Thisreportpresentsthetechnologiesandsystemsevaluatedanddetailstheresultsofa

simulationstudydesignedtoestimatethetotalsysteminstallmentandoperatingcostversus

detectionperformanceforthesesystems.Theanalysiswasnotintendedtoprovide

comprehensivecosteffectivenessestimatesofthesesystems,anddoesnotincludeactions,costs,andoutcomesfollowingdetectionofadiseaseoutbreak.Theresultshelpinformdecisionmakersinterestedindeployingsuchasysteminaspecificlocationofthecoststheywouldexpectto

incur,andhowthesecostsrelatetodetectionoutcomes.

ThreeTechnologiesAssessedinThisStudy

Wedecidedtofocusoncomplementarytechnologieswhichcanbedeployedincombinationfornovelpathogendetection.Thesetechnologieswerechosenduetohavingoverlappingbut

distinctcharacteristics(cost,easeofdeployment,technologyreadiness,etc),collectivelyintendedtosupport“pathogenagnostic”workflows.

Thethreetechnologieschosenwerewearablesphysiologicalsensors(Wearables),multiplexPolymeraseChainReaction(mPCR),andmetagenomicNext-GenerationSequencing(mNGS).Respectively,thesetechnologiesprovidei)continuoussyndromicmonitoringofentire

populations,ii)broadbutpathogen-specificmoleculardetection,andiii)pathogenagnosticdetectionandrichthreatcharacterization.Eachisdescribedinmoredetailbelow.

WearablePhysiologicalSensors

Wearablesarecontinuouslywornphysiologicalsensorsthatcancomplementtraditional

diseasesurveillancesystemsbycontinuouslymonitoringphysiologicalparameterssuchasheartrate,bodytemperature,sleeppatterns,andbloodoxygenlevels(Kimetal.,2019).Subtle

3

deviationsfromnormalpatternscouldindicatetheonsetofaninfectionbeforephysicalsymptomsmanifest,andwellbeforepresentationatahealthcarefacility.

MultiplexPolymeraseChainReaction

mPCRpanelsscreenasinglesampleagainstavarietyofknownthreats,allowingforbroaddiagnosesundirectedbythepatient’ssymptoms(Wittweretal.,2001).Forourpurposesinthisanalysis,mPCRpanelsallow“rulingout”ofknown,commoncausesofillness,whichreduces

thenumberofsamplesreferredforwardforpathogen-agnosticmetagenomicsequencing,andhastheaddedadvantageofnotrequiringsignificantmarginalequipmentcosts.

MetagenomicSequencing

mNGScanbeusedtodetectanovelpathogenwithoutpriorknowledgeofitsgenomeand

demonstratessignificantpromiseinitsutilityasapathogen-agnosticsurveillancetool.mNGScouldalsoreducethetimebetweennovelpathogendiscoveryandgeneticcharacterization,

whichmayacceleratethedevelopmentanddeploymentofcountermeasuresthatrequiregenomeinformation.TherehavebeencallsforincorporatingmNGSintoglobalsurveillancefornew

outbreakdetectionandreducingthetimefromdiscoverytopathogencharacterization(Chiu&Miller,2019;Gardy&Loman,2018;Milleretal.,2013).

Methods

Ourbase-casescenariosimulatedanoutbreakofapreviouslyunknownairbornepathogensimilartoSARS-Cov-2spreadingperson-to-personinaclosedpopulation,beginningwitha

singleinfectedindividual.Thisscenariorepresentsanovelrespiratorypathogenwithpandemicpotential,spreadingthroughapopulationofdeployedmilitarypersonnel.

TechnicaldetailsofourepidemicsimulationcanbefoundinAppendixB(seeTransmissionModelsection,andAppendixBTable1).Briefly,weassumeeachhealthcarefacilityserveda

closedpopulationof1000susceptibleindividuals.Transmissionisexclusivelyperson-to-personwithapre-symptomaticperiodduringwhichaninfectedpersoncaninfectothers.Further,someinfectedindividualscaninfectotherswhileremainingasymptomatic.Baselineestimatesfor

parametersinthismodelreflectingcharacteristicsofthedisease,population,andtechnologies

wereidentifiedfromliteratureandexpertengagement.Therangeofvaluesused,nominalvalues,andtheirsourcesarepresentedinAppendixB.

ThenominalvalueofeachparameterismeanttorepresentapandemiccausedbyapathogensimilartowildtypeSARS-Cov-2,whereassensitivityanalysesevaluatetheperformanceofthesystemsacrossawidesetofscenariosrepresentingpathogenswithdifferenttransmissibilityandalternativeassumptionsabouttechnologyperformance.

Weassessedthreesurveillancestrategies,eachofwhichcombinedtraditionalandnoveldetectiontechnologiesinalogicalorder.Followingstandardpracticetoincreasecost-

4

effectiveness,weusedlowercost-per-usetechnologyforinitialscreeningandreservedhighercost-per-usetechnologyatalaterstageforsamplesthatweremorelikelytocontainanovel

pathogen.Thiswasintendedtomakeeachstrategyindependentlyascost-effectiveaspossiblewhileallowingcomparisonsofglobalcost-effectivenessacrossstrategiesforarangeofthreatscenarios,plausiblevaluesforuncertainquantities,andotherimplementationchoices.

Threetypesofcosts—fixedupfrontcostofinstallation,ongoingoperatingcost,andper-testvariablecost—wereestimatedthroughi)peer-reviewedliteratureincludingpublisheddatafromprioranalyses,ii)consultationswiththeauthorsofthesestudies,andiii)publiclyavailable

informationfrommanufacturers.DetailsoncostsandsourcesareprovidedinAppendixB.

SurveillanceStrategies

Toevaluatecost-effectivenessinarealisticdeploymentsetting,wheredetectiontechnologiesareunlikelytobeusedinisolation,weconsideredthree“strategies”fordetectingnovel

pathogens.Thesearebrieflydescribedbelow,anddetaileddescriptionsofeachstrategyare

providedinAppendixA.Theproportionofindividualsreferredtothenextstageofeachstrategywasvariedthrougharangeofplausiblevalues,asdetailedinAppendixB.

Syndromic

AfractionofsymptomaticindividualswhopresenttoahealthcarefacilityarefirstinvestigatedbyanmPCRpanel,withaproportionoftheremainingunexplainedsymptomaticindividuals

referredformNGS.TheSyndromicstrategymostcloselymirrorscurrentlyusedbiosurveillanceearlywarningsystems,withtheadditionofsystematicthreatcharacterizationthroughpathogen-agnosticmNGStesting.

Wearable

Similarly,inthisstrategyafractionofindividualswhoreceiveanalertfromapassivewearablesensorarefirstscreenedagainstknownrespiratoryillnesscausesusinganmPCRpanel,anda

fractionofstill-unexplainedalertsarethenfurtherinvestigatedviasystematicpathogen-agnosticmNGStesting.

Environmental1

Inthisstrategy,sequencingofbothwastewatersamplesandswabsfromarandomsubsetofthepopulationoccursatabaselinelowfrequency,whichcanbeintensified(intermsoffrequency,

1NotethatwhiletheSyndromicandWearablestrategiesarestructurallysimilarandcanbedirectlycompared,theEnvironmentalstrategyincludestheadditionalelementofadaptiveintensificationinresponsetooutside

information.ThissomewhatlimitsthedirectcomparabilityoftheEnvironmentalstrategy,andfutureworkcouldexploretheconsequencesofincludingthiselementintheotherstrategiesorassesstheperformanceofan

Environmentalstrategywithoutthisfeature.Further,sincethisstrategydoesnottestbasedonsymptomologyitismorevulnerabletofalsepositiveresults.Forexample,itispossiblethatbothwastewatersitesandindividual

5

populationfraction,and/orsequencingdepth)inresponsetoanexternalsignalorwarning(e.g.fromanevent-basedsurveillancesystem).Concurrentsamplingofwastewaterandswabsallowsforcross-comparisonofresultsfromacombinedcommunitysource(i.e.,wastewater)and

individualsinthepopulation,allowingforcorroborationthatanovelsequencedetectedinwastewatercorrespondstoanovelhumanpathogen.

SurveillanceCosts

Weestimatedcapital,operating,andper-testexpendituresassociatedwitheachstrategy.Forallstrategiesweassumedthepre-existenceandavailabilityoftypicallaboratorycapabilitiesandthusdidnotincludeadditionalcostsforroutinematerialsandequipment(includingPCR

machinesandstandardrelatedconsumables).Capitalcostsincludedinvestmentsinsequencingmachines,wearabledevices,andpersonalprotectiveequipment(specifically,2protectivere-

usablesuitsforwastewatersamplingintheEnvironmentalstrategy,perfacility),whileoperatingcostsincludedlaborandfacilityexpenses.Variableper-testcostsassociatedwithmPCRand

mNGSincludedmPCRtestkits,sequencingpreparationandreagents,aswellasothernon-

standardconsumables.Thesecostestimateswerebasedonprevailingmarketratesfor

comparablegoodsandstandardlaboratoryprocedures.AmoredetaileddescriptionofthecostsanddatasourcesisprovidedinAppendixB(seeTablesB.5toB.7fordetailedcostinformation).

ScenarioDiscovery

Apriori,weexpecttheSyndromicsurveillancesystemtounderperformindetectionrelativetotheWearableandEnvironmentalStrategies.Yetitisuncleartheextenttowhichthisresult

willholdacrossawidesetofpotentialnewpathogensandtechnologyperformance

characteristics.Therefore,weuseaScenarioDiscoveryapproach(Bryant&Lempert,2010)toidentifyconditionsunderwhicheachstrategywouldfailtodetecttheoutbreak15daysbeforethesyndromicstrategywouldotherwisedetectit.Wefocusontimetodetectionasthekey

outcomeinthisanalysisbecauseearlydetectionisthemaingoalofthosesystems.

WeconstructaLatinhypercubesample(LHS)encompassingtheplausiblerangeofall

parameterstoinvestigateconditionsunderwhichsyndromicsurveillancedetectsoutbreaks

withincomparabletimeframestotheotherstrategies.Specifically,wefirstrunthethree

strategiesinabase-casescenarioaimedtorepresentthecharacteristicsofapathogensimilarlytoSARS-Cov-2.Becauseeachstrategyhasadifferentsetofparameters(TablesB2,B3andB4),itisinefficienttocreateonesamplevaryingallparameterssimultaneously.Therefore,wecreate

twosetsofparameters:onetocomparetheWearablesstrategytotheSyndromicstrategy

(parametersinTablesB1,B2andB3)andonetocomparetheEnvironmentalStrategytotheSyndromicstrategy(parametersinTablesB1,B2andB4).Then,foreachparameterset,we

samplescouldbecontaminatedbyanovelnon-humanvirus(e.g.acircovirus),whichcouldproduceafalsepositivesignalthatwouldrequirefurtherinvestigationtoascertainasnotassociatedwithhumandisease.

6

comparethetime-to-detectionoftheWearableortheEnvironmentalstrategytotheSyndromicsurveillancestrategy,computingthenumberofdaysof“earlywarning”ofeachstrategyrelativetoSyndromicsurveillance.Inotherwords,wecomputea“detectiontimelinessregret”—i.e.,thenumberofadditionaldaysthatitwouldtaketodetectapathogenifthesyndromicstrategywasusedinlieuoftheothertwoalternatives.

Wethenapplyscenariodiscoverymethodstocharacterize“latewarning”scenarios—i.e.,

conditionsunderwhichenvironmentalorwearablestrategiesfailedtodetecttheoutbreak15

daysbeforesyndromicsurveillance.Our“latewarning”scenariothereforeiscodedasabinaryoutcome:scenarioswheresyndromicsurveillancedetectswithin15daysofthealternative

strategyarelabeledas“latewarning”cases,whilescenarioswherethealternativedetectsmorethan15daysbeforesyndromicarecodedas“earlywarning”cases.Theindependentvariablescomprise27parametersspanningpathogencharacteristics(reproductionnumber,asymptomaticprevalence,latentandrecoveryperiodrates),surveillanceimplementationfactors(metagenomicsequencingprobabilities,clinicvisitrates,testingsensitivities),andtechnicalspecifications

(wastewaterdetectionthresholds,sequencingdepth,sensorperformance).

UsingthePatientRuleInductionMethod(PRIM),wecharacterizeparameterregionswheresyndromicdetectionoccurswithin15daysofalternativestrategies.Inaddition,wealsouse

ClassificationandRegressionTrees(CART)toidentifyregionsintheparameterspac