2024對(duì)話式AI技術(shù)和遠(yuǎn)程患者監(jiān)測(cè)（RPM）白皮書(shū)

CONVERSATIONALAITECHNOLOGYAND

REMOTEPATIENT

MONITORING(RPM)

ACALLTOACTION

Forthelatest,detailedinformationontheworkoftheOpenVoice

projectsoftheLinuxAI&DataFoundation,pleasevisitourwebsiteat

andtheOpenVoiceNetworkGitHubRepositoryat

/open-voice-trustmark/docs

Version2.0

December2023

ExecutiveSummary

ThisreportbytheHealth,Wellness,andLifeSciences

CommunityoftheOpenVoiceTrustMarkInitiativeoftheLFAI&DataFoundationwascreatedforhealthcareprofessionals,organizations,andenterprisesconsidering,leveraging,or

implementingconversationalartificialintelligencetechnology(includingvoiceassistanceandgenerativeAIapplications)inRemotePatientMonitoring(RPM)/telehealthapplications.

?Asindustryexpertsknowalltoowell,thereisagrowing

crisisinhealthandwellnessdelivery,drivenbydemographic,professional,andsocio-economicmarkettransitions.

Technology-enabledRPMisanimportantresponsetothiscrisis.NewresearchbyJuniperResearch(RemotePatientMonitoring2023)forecastsan8.8%CAGRinRPMadoption2023-2027,toaglobalinvestmentof$110billion.

?RPMisavitalcomponentoftelehealth.ThispaperarguesthatconversationalAItechnologies–mostimportantly,voiceassistance–canandwillbeanimportantpartofRPM.Infact,earlyimplementationssuggestthatacrossnumerousclinicalusecases,conditions,andsymptomsconversationalAItechnologiesandanalysisdrive

betterpatientoutcomeswithsignificantlyenhancedimplementationandclinicalefficiencies.

?TherearemanyinstancesinwhichconversationalAI

technologiesarenowenablingeffectiveRPMandbetter

patientoutcomes.Biomarkeranalysis–theidentificationofleadingindicatorsofmentalandphysicalconditions–mayofferthegreatestvalueintheprovisionofcontinuouscare.

ThispapernotesseveralexamplesofRPMconversationalAIuse.

?IntheUnitedStates,pathsofreimbursementforremote

patientcaregrewwiththefirstimplementationsof

telehealth,andacceleratedduringtheglobal2020-2022

pandemic,asmeasuredbyadvancesinCurrentProceduralTerminology(CPT)codingandCenterforMedicare/MedicaidServices(CMS)policies.However,asofpublication,FederalreimbursementforRPMcontinuestoexpand,butstate

Medicaidreimbursementremainsapatchworkquilt,

andrecognitionorclearanceofSoftwareasaMedical

Device(SaMD)–softwarebeingthecoreofconversationalassistance–isstillanoutstandingissue.

?VoiceandconversationalassistanceAItechnologiesbringwiththemimportantlegal,ethical,andtechnological

issuesthatbegin–butdonotend–withprivacy.ThispaperandtheTrustMarkInitiativeprojectoftheLinuxAI&DataFoundationprovidesguidanceforriskmitigation.

CONVERSATIONALAITECHNOLOGYANDREMOTEPATIENTMONITORING(RPM)|LFAI&DATA2

Introduction

WeareatthebeginningofanewchapterinRemotePatient

Monitoring(RPM)–oneincreasinglyaudiblethrough

conversationalassistancetechnology,thecombinationof

naturallanguageprocessing,understanding,andgeneration

(oftentermed“voice”)andnaturallanguagegenerativeartificialintelligence.

RPMisnothingnew.Industryhistorianspointtothe1948

transmissionofradiologyimagesovertelephonelines,theearly

1960’smonitoringofbloodpressuredatafromMercurymission

astronauts,andthe1970’sRPMprogramattheArizona(U.S.)

PapagoIndianReservationasseminalmoments.Inthe21st

century,technologyandindustryadvancements–fromwireless

internet(Wi-Fi),theshort-rangecommunicationtechnologyknownasBluetooth,broadband,andchangestoCPTcodes–andthe

COVID-19globalpandemic–ledtorapidgrowthintheuseofRPMtechnology(andexpansionofRPMcoveragebygovernmentandprivatepayers).

Thispaper–withrespecttothetechnologiesandprocessesthathaveblazedRPM’strailtodate–pointstotheneedforindustry-widerecognitionthatconversationalAItechnologycanandwillplayanimportantroleinthefutureofRPMworldwide.

Ourthesis:ConversationalAI–thecombinationof

artificialintelligence-enablednaturallanguageprocessing,understanding,generation,anddataanalysis,alongwithnewtoolsofgenerativeAI–canprovidepatients,clinicians,andprovidersRPMatlowercostwithincreasedadoption,broadutility,andexpansiveinclusivity–andever-preciseidentificationofleadingindicatorsofalengtheninglistofphysicaldiseasesandmentalhealthdisorders.

Giventhechallengeofagingandunderservedpopulations,a

shortageofclinicians,andagrowinghealthequitygap,itbehoovesclinicians,providers,payers,andgovernmentstoexplore,test,–

andrapidlyexpand–theuseofandreimbursementforRPMusingthetoolsofconversationalAI.

Despiteitsmanyimplementationsandsuccesses,RPMisstillin

itsearlydays.TheindustryissuesthatledtothecurrentstateofRPMareexpandingandaccelerating.Expertsforecastashortageofphysiciansandhealthcareworkersthatmaygrowto15millionworldwideby2030.Acrossdevelopedeconomies,thepopulationisaging–andisincreasinglysaddledwithlife-shorteningchronicconditions.IntheUnitedStates,accordingtogovernmentsources,some10,000individualsoftheBabyBoomgenerationnowreachtheir65thbirthdayeachday(U.S.HHS.com).Inimpoverishedandruralcommunities,clinicalfacilitiesareincreasinglyscarce.

And–despitegrowingrecognitionofandappreciationfor

RPM,aMedicalGroupManagementAssociationpollrevealedthatjust25%ofU.S.practicesofferRPMservices.

CONVERSATIONALAITECHNOLOGYANDREMOTEPATIENTMONITORING(RPM)|LFAI&DATA3

TableofContents

Situation:AnEver-WideningCrisis

5

RemotePatientMonitoring:AnEvolvingSolutiontotheGrowingProblem

6

TheCapabilitiesAddedtoRPMbyConversationalAI/VoiceTechnologies

7

TheEconomicandOperationalValueAddedtoRPMbyConversationalAI/VoiceTechnologies

8

Today’sRPMVoiceTechnologyToolkit

9

ForDevelopers:a0.1ListofConversationalAIRPMRequirements

9

ThreeSignificantChallengestoConversationalAIinRPM

10

Reimbursement

10

GeneralGuidanceforTechnologyDevelopersRegardingBillingandReimbursement

11

GenerativeAI,LargeLanguageModels,andHealthCare

11

LawandEthics:EarningTrust,PromotingDignity

12

CoreEthicalPrinciplesforConversationalAI

12

TheUniqueValueandRisksofVoiceData

13

InclusiveEngagementtoReachMorePatientPopulations

14

AccessibleEngagement

15

VoiceandConversationalAIinRPM:CurrentImplementationsandTrials

15

TheVanderbilt“HospitalatHome”P(pán)rogram:RPMforShort-TermTreatment

15

WoltersKluwerHealth:VoiceforEducation,BehaviorChange,andReducingReadmission

16

Emids:AHolisticConversationalCoachtoCuttheSpecialist’sWaitlist

17

EU-JapanProjectUsesConversationalAIforVirtualSeniorEngagementandCompanionship

17

Orlov:VoiceAssistancetoMitigateSocialIsolation,EnableSecureAccesstoHealthInformation

17

VoiceTechnologyasaComplementtoWearablesandOtherRPMTechnologies

18

BiomarkerAnalysis:FindingIndicatorsofPhysical,MentalHealthinVoice

18

TQIntelligence:VoiceforChildandAdolescentMentalHealth

18

BiomarkerAnalysisforPrecisionMedicineSupport

19

VoiceforEarlyDetectionofRespiratoryDisease

19

SupportforVoiceBiomarkerResearchExpandstheMedicalPossibilitiesforRPM

20

Public/PrivatePartnershipsDesignedtoAccelerateVocalBiomarkerDatabaseGrowth

20

PastandRecentInvestmentsinVoiceBiomarkerResearch

21

Conclusion

22

Acknowledgements

23

AbouttheOpenVoiceTrustMarkInitiativeoftheLFAI&DataFoundation

23

AbouttheLFAI&DataFoundation

23

AboutTheLinuxFoundation

23

AppendixA.VocabularyTerms

24

VoiceTechnology

24

ReferenceList

24

CONVERSATIONALAITECHNOLOGYANDREMOTEPATIENTMONITORING(RPM)|LFAI&DATA4

Situation:AnEver-WideningCrisis

ThesearesomeofthereasonswhyRPM–specifically,RPMof

greateradoption,greateruse,greaterinclusivity,andlowercost–issoimportant:

?ThereisashortageofphysiciansintheUnitedStates.

?A2021reportpublishedbytheAssociationofAmericanMedicalCollegesprojectstheU.S.couldseeanestimatedshortageofbetween37,800and124,000physiciansby

2034,includingshortfallsinbothprimaryandspecialtycare(AssociationofAmericanMedicalColleges,2021).

?Thereisashortageofhealthcareworkersworldwide.

?TheWorldHealthOrganization(WHO)estimatesthat

therewillbeashortageof15millionhealthprofessionalsinhospitals,nursinghomes,andcarecentersworldwideby2030(Stroobants,J.etal.,2022).

?Thereisadisparityinhealthequity–especiallyforunderservedpopulations.

?“5VulnerablePopulationsinHealthcare”publishedintheAmericanJournalofManagedCare(Joszt2018)citesthefollowingasunderservedpopulations:

?Chronicallyillanddisabled;

?Low-incomeand/orhouselessindividuals;

?Certaingeographicalcommunities,suchasruralareasoftheUSandNativeAmericanslivingonreservations;

?LGBTQ+population

?Theveryyoungandveryold.

?Therearenumerousmedicallyunderservedareas/populations(MUA/P)intheUS.

?MUA/ParedesignatedbytheHealthResources&

ServiceAdministrationashavinganinsufficientnumberofprimarycareproviders,highinfantmortality,high

poverty,orahighelderlypopulation.AreportpublishedbytheUniversityofMedicineandHealthSciencescitesfourmaincommunitiesinneedofdoctors(Harrah2020):

?MigrantandHispanicpopulations

?Ruralcommunities

?AfricanAmericaninner-cityareas

?AsiaAmericanandPacificIslanderinner-cityareas.

?Thedevelopedworldpopulationisaging.

?InOctober2022,theWorldHealthOrganizationreportedthatbetween2015and2050theproportionofthe

world’spopulationover60yearswillnearlydoublefrom12%to22%(WorldHealthOrganization,2022).

?Thereisanincreaseinchronicconditions.

?IntheU.S.alone,86%ofhealthcarecostsaregoing

towardanepidemicofchronichealthconditions(Holman2020).

?Thereisadigitaldividewithindevelopedworld

populations:thosewith–andcomfortablewith–

technology,andthosewithoutit.IntheUnitedStatesasof2021(L),

?only77percentofhomeshadbroadbandaccess–

anumberthatdroppedto65percentofHispanic

households,andto57percentinhouseholdswith

householdincome(HHI)of$30,000orless.In

comparison,broadbandwaspresentin92percentofUShomeswithhouseholdincomeof$75,000ormore.

CONVERSATIONALAITECHNOLOGYANDREMOTEPATIENTMONITORING(RPM)|LFAI&DATA5

RemotePatientMonitoring:AnEvolvingSolutiontotheGrowingProblem

TheAmericanTelemedicineAssociationRemoteMonitoringSpecialInterestGroup(RMSIG),definesRPMthisway:

thecollectionofclinicallyrelevantinformationaboutan

individualwhichisreviewedbyacareteamand/orcare

processes.Theclinicianandthepatient/clientarenotin

thesameroom.Medicaldevicesmayormaynotbeutilized.Datacanbeself-reportedor/ordigitallyuploadedbythe

patientoranon-siteaide.(

/

community/remote-monitoring/

)

TherootsofRemotePatientMonitoringcanbetracedtoa

1948Philadelphia-areatransmissionofradiologyimages

overatelephoneline,and1960psychiatricconsultations

deliveredthroughclosed-circuittelevision.Otherimportant

earlymilestonesinRPMdevelopmentcamein1961,astheU.S.

NationalAeronauticsandSpaceAdministration(NASA)monitoredthein-flighthealthofastronautAlanShepardwithearlyEKG

technology,athermometer,andarespirationsensorinhis

microphone–afirstuseofvoiceinRPM;thedevelopmentanddeployment,inthe1970’s,ofanRPMprogramforthepopulationofthePapagoReservationinArizona;andin1982,withthe

introductionofthefirstwearablebiometricsportswatch–onethatusedEKGtechnologytomeasureheartrateandactivity.(Paul/P,February2023).

Inmorerecentyears,theRPMopportunityhasbeenenhancedbytheturn-of-the-21stcenturyintroductionofthewirelessinternet(Wi-Fi),short-rangecommunicationtechnology(Bluetooth),andthespreadandadoptionofever-fasterbroadbandandcellular

networks.

Thesetechnologicalinnovationshaveenabledcliniciansand

providerstoconnectwithmorepatientpopulations–andtodosowithlesscost,greaterspeed,andwithmorediagnosticprecision.Overall,RPMpromisesto

?DeliverValue-BasedCare(VBC)andFee-for-Service

(FFS)costmanagement,byreducingunscheduledacute-carecosts,especiallythroughthereductionofhospital

admissions,readmissions,andemergencyroomvisits

?Helpovercomeclinicianshortages,especiallyamongruralandotherunderservedpopulations

?Expandmedicalcaretopopulationsunderservedforreasonsofgeography,economics,culture,andlanguage

?Improveaccessibilitytomentalhealthservices

?Reducecancelledappointmentsandno-shows

?Helphourly-wagepatientsavoidlossofincome

?Enableolderanddisabledindividualstoliveathomelongerandavoidordelaytransitionintomoreexpensiveskillednursingfacilities

?Providealevelofcontinuous,value-basedcare,whichpromisestoreducethenumberofhospitalizations,

readmissions,andlengthsofstayinhospital–allofwhichhelpimprovequalityoflifeandcontaincosts(CenterforConnectedHealthPolicy)

?Effectivelyandefficientlyscreenpopulationsintimesofinfectiousdisease/pandemics

?Extendaccesstocarebeyondnormalhours

?Reducepatientandprovidertravelburden

TheuseofRPMisgrowingrapidly.NewresearchbyJuniper

Research(RemotePatientMonitoring2023)forecaststheadoptionofRPMwithinbothdevelopingmarketsandestablishedmarketstobe$72billionin2023and$110billionin2027,with2023-2027

growthat54%.

CONVERSATIONALAITECHNOLOGYANDREMOTEPATIENTMONITORING(RPM)|LFAI&DATA6

Infact,RPM’svaluetopatients,clinicians,andprovidersis

reflectedinapushfornewrules,suchastheU.S.CentersforMedicareandMedicaidServices’(CMS)changestohandlingRemoteTherapeuticMonitoringreimbursementandthe

additionofCognitiveBehavioralTherapyintheFinalRule

PhysicianFeeSchedulefor2024.NixonGwiltLaw(NGL),whichspecializesinhealthcareandinnovation,notesthey“havemaderecommendationstorevisittheoverallcodestructuretobetteralignRTMwithRemotePhysiologicMonitoringandothercare

managementservices”(O’Connor,K.etal,2022).

In2023,HomeHealthAgenciesbeganmandatoryreportingof

RPM,utilizingcodeG0322:thecollectionofphysiologicdata

digitallystoredand/ortransmittedbythepatienttothehome

healthagency(forexample,remotepatientmonitoring),andin

2024FederallyQualifiedHealthCenters(FQHCs)willbeabletobillforRPMusingcode(G0511).(Paschall,2023).

WorkingDefinitionofConversationalAI

ConversationalAIimplementsvoicetechnology

inaudibleand/ortextualautomatedinterchanges

betweenahumananddigitaldevice–oftenthroughvirtualvoiceassistancesoftware.SuchinterchangescanfacilitatesInteractive,automatedpatient

engagement;inclusiveengagement;accessibleengagement;andbiomarkerprecisionmedicine.

TheCapabilitiesAddedtoRPMby

ConversationalAI/VoiceTechnologies

ConversationalAItechnologies–andspecifically,voice

technologies–areanewandgame-changingadditiontotheRPMmix.

RemotePatientMonitoringrequiresbiometricdatatransfer,

andmanyRPMplatformsincludetelehealth(real-timeaudio

and/orvideo.)Whenpaired,telehealthprovidestheepisodic

communicationandcare,andRPMprovidescontinualbiometric(objective)datatosurveyresponse(subjective)datatocompletetheremotecarecycle.(Paschall,2023)

Thatdefinitioncannowevolve,thankstorecentvoicetechnologyresearchandinnovation.Broadlydefined,voicetechnologiesofferthefollowing,throughsoftwarethatcanresideonmultipletypesofdevices:

?naturallanguageunderstandingandprocessing–theabilityofacomputationaldeviceto“l(fā)isten”tohuman

utterancesandtranslatespokenwordsintotext,acrossdozensoflanguagesanddialects

?naturallanguagegeneration–theabilityofa

computationaldevicestoturntextintohuman-likespeechacrossdozensoflanguagesanddialects

?voiceassistance–throughwhichacomputercan

proactivelyinteractwithahuman(e.g.,providingreminders,aclinicalinterview,orentertainment)orrespondtohumaninquirieswithtrustworthyinformationona24/7basis.Thiscanenableinteractive,automatedpatientengagement,

including

?briefverbaldiaryentries

?conversationalappswithpatient/caregiverinstructionsandvirtualcoaches

?data-gathering/reportingtoolsthatusevoicetechnologytogatherandreportnewdataandinteractwithwearabledevicesthattrackfitnessandhealthconditionsthroughbloodpressure,glucose,heartrate,anddigitalECG/EKGmonitors

?anautomatedmeansto

?obtainpatientconsent?confirmpatienteligibility

?on-boardpatientsintocareprocesses

CONVERSATIONALAITECHNOLOGYANDREMOTEPATIENTMONITORING(RPM)|LFAI&DATA7

?ameansforcontinuouscareandevaluationofchronicconditions,providing:

?Informationonleadinghealthindicators?automatedreminders

?instruction/education

?caregivercoordinationanddecisionsupport?companionship

?Atrustworthysourceof24/7information

?inclusiveandaccessiblepatientengagementacrossmultiplelanguages,dialects,speechdifferences,andunderservedpopulations(PewResearchCenter,2021)

?AI-basedspeech/voiceanalysis,whichcanprovide

assessmentsofleadingindicatorsofthespeaker’smentalandphysicalhealth,aswellasidentify(withincreasing

accuracy)thespeaker’sgender,geographicregion,ethnicity,educational-incomelevel,bodymassindex,upperbody

strength,andpersonalitytype,andemotion.Thehumanvoiceis:

?abiometricidentifier,similartofingerprintsintheabilitytoidentifyauniqueindividual

?abiomarker(definedbyOxforda

measurable

substanceinan

organism

whosepresenceisindicativeofsomephenomenonsuchasdisease,infection,or

environmentalexposure),which,invoice,canprovideincreasinglypreciseindicatorsofmentalandphysicalillness,rangingfromschizophreniatoParkinson’sandrespiratoryconditionssuchasCOVID-19.

TheEconomicandOperationalValueAddedtoRPMbyConversationalAI/VoiceTechnologies

ConversationalAItechnologiesaddthefollowingtothecorevaluepropositionofRPMnotedabove:

?Ofnote:thewidespreadadoptionofsmartphonesacrossallagesandincomelevelsofferssignificant

potentialforconversational/voice-enabledRPM.In

2023,91percentofAmericansownedandoperatedsmartphones(EdisonResearch,2023);in2021,

smartphoneswereownedby76percentofAmericanadultswithhouseholdincomesoflessthan$30,000–apercentagesignificantlyhigherthanhouseholdsofsimilarincomewithbroadband(seeabove).

?Smartphoneownersincreasinglyusetheirphonestomonitorhealth.Deloitte’s2022ConnectivityandMobileTrendsSurvey(DeloitteCenterforTechnology,Media&Telecommunications,2022)reported:

?Atleastathirdofsmartphoneusersaremonitoringtheirhealthandfitnesswiththeirphones,and1in5usemeditationormentalwellnessapps.

?Ninein10consumerswhoownthesedevicesusethemtotrackfitnessandmonitorhealthmetrics.

?Voiceisdeliveredthroughsoftware,whichcanbe

installedonmultipledevices–especiallysmartphones.The2021U.S.installedbaseofvoiceassistancesoftwareonsmartphones(140.1million)exceededby54percentthatofsmartspeakers(90.7million)(voicebot.ai,2022).

?RPMvoiceservicescanbedeliveredaccurately

throughcellulartechnology.Homebroadbandand/orWi-Fimaynotberequired.

?Voicetechnologyisalsoaregularelementwithinsmarthomes,andisincreasinglyapartofremotecontrolsand

smartappliances

.

?Lowerbarrierstoadoption:asnotedabove,ahigh

percentageofprospectivepatientsmayalreadyownandusedailythedigitaltoolsrequiredforconversationalAI-based

RPMservices

?Lowerimplementationandoperatingcosts:manyconversationalAI-RPMservicescanbeprovidedusingexistingandwidelyadoptedconsumertechnologies–smartphones,cellularcommunication

CONVERSATIONALAITECHNOLOGYANDREMOTEPATIENTMONITORING(RPM)|LFAI&DATA8

?Inclusivity:theabilityofconversationalAItechnologiestolistenandspeakinthelanguageoffluencyforbothpatientandprovider(andtotranslateonelanguagetotheother,whennecessary)

?Biomarkeranalysis:theabilitytodetectinapatient’svoiceleadingindicatorsofanxiety,stress,andmentalandphysicalillness

?Increasedautomationofclinician/providerdailywork:

theability,throughnaturallanguagespeech-to-text

understanding,toprovidetranscripts(inmultiplelanguagesanddialects)forrecordkeeping,thesubmissionofclaims,andcommunicationwithpatients.

?Forclinicians/providers,conversationalAIprovidestheabilitytocapturepatient-clinicianinteractioninbothdirectandautomatedconnectionwithremotepatients.This

enables:

?Acousticanalysisandsentimentanalysisofvoice

biomarkerstomonitorkeyhealthindicatorsfor

awareness/diagnosisandcontinuouslyevaluatechronicconditions

?Hands-freeclinicalcare

Today’sRPMVoiceTechnologyToolkit

Today’sRPMvoicetechnologytoolkitincludes

?Voicebotsthatcan:

?automaticallyandaccuratelytranscribeclinicalconversations

?provideinformation24/7thataccuratelyreflects

scientificandclinicianconsensusregardingsymptomsandconditions

?listen,translate,andspeakindozensoflanguages

?engagesolitarypatientswithreminders,alerts,communication,andentertainment

?scheduleandmanageappointments

?beaccessedthroughcommonlyusedpersonaltechnology,suchasthesmartphone

?beusedfordirect,real-timeinteractionwithclinicians?textualandacousticanalysisofwhatissaid,whichcan:

?gaugepatientsentiment,includinganxietyandstress

?provideever-precisebiomarkeranalysisofmentalandphysicalillnessesrangingfromschizophreniatoParkinson’sandCOVID-19

Thesecapabilitiescanprovidepatient,clinician,andprovidervaluebythemselves,orasacomplementtootherRPMapproacheswithwearable,at-home,ortelehealthtechnologies.

ForDevelopers:a0.1ListofConversationalAIRPMRequirements

ThefollowingissuggestedasastartingpointforConversationalAIRPMdevelopers.Thislistwillbeextendedandrevisedovertime.

AConversationalAIRPMimplementation–byitself,oraspartofalargerRPMsystem–MUSTprovide:

?AcceptancewithinCurrentProceduralTerminology(CPT)codesandCenterforMedicare/MedicaidServices(CMS)billing(seebelow)

?Processesthatallowpatientstoscheduleappointments

?Processesforclearcommunicationandacquisitionofpatientconsent

?Unidirectionalorbidirectionalintegrationswithelectronicmedicalrecords

?Accuratetimetrackingofinteractionanduseforthepurposeofreimbursement

?Carepathwaysthatarecustomizablebysymptomologyandpatient

CONVERSATIONALAITECHNOLOGYANDREMOTEPATIENTMONITORING(RPM)|LFAI&DATA9

?Arangeofmodalitiesforengagingpatientsaccordingto

patientconditionandenvironment,andaccordingtothe

technologyavailability,e.g.,broadband,WI-FI,cellularwithintheenvironment

?Single,securesign-on

?Continuouscareservices,whichcouldincludedaily

orperiodicbiometricassessment,surveyquestions,

educationalcontent,andvideoconferencing,aswellastheprovisionofinterventionalertstopractitioners

?Naturallanguageinclusivityservices,whichenablepatients

tobeaccuratelyheardandspokentointheirlanguageordialectoffluency

?ITsupportforpatientandprovider

?Documentedpoliciesastotheprivacyandclinicaluseofpatientvoicedata(whetheracousticortranslatedtotext)

?Datasecuritythatprovidespatientandprovidersafety.(“TelehealthforProviders:WhatYouNeedtoKnow,”U.S.

DepartmentofHealthandHumanServices,2023,May;Bill

Paschall,ClearArchHealth,presentationtotheOpenVoice

NetworkHealth,Wellness,andLifeSciencesCommunity,2023,

June).

ThreeSignificantChallengestoConversationalAIinRPM

ThedevelopmentofRPMsolutionsusingconversationalAI

technologiesmustincludecarefulconsiderationofthree

significantissues:1.reimbursement,2.theuseoflargelanguagemodels,and3.thelegal/ethicaluseofvoicedata.

Reimbursement

BillingandreimbursementforRPMisanevolving,oftenpatchworkissue.BillingandreimbursementrequirementsforRPMservices–forphysicalandmentalhealth--varyamongdifferentpayers/

insurersandfordifferentgeographiclocations(U.S.HHS,May2023).

Today’spolicies,regulations,andrequirementshaveevolved

withtechnologicalchangeoverthepastdecade.TheCOVID-19

globalpandemicledtotime-boundacceptanceofselectedRPMpractices;insomecases,acceptancehasbeenextendedormadepermanent–insomecases,not.Forinstance,asofQ42023,

Medicarereimbursementforaudio-onlytelehealthservices(a

COVID-19innovation)iscurrentlyonlycoveredthroughDecember31,2024.

Thispaper,inaccordwithU.S.DepartmentofHealthandHumanServicesguidancepublishedearlierthisyear,willnotattempttoguidereadersonbillingandreimbursementdetailsandpractices.Providersarestronglyencouragedtoconnectwithbillingand

reimbursementexpertstoclearlyunderstanddefinitions,requirements,andprocesses.

Thetechnologycommunity–especiallythoseenvisioningor

pursuingRPMsolutionswithandthroughconversationalAI–

mustopenanddriveconversationswiththeCenterforMedicare/MedicaidServicesastowhatconversationalAIcando,whatitis

appropriatefor,andhowitcanreplace/supplementhuman-to-

humancommunicationacrossthetelehealth/RPMreimbursementcodes.

Dr.EmreSezgin,whohasservedasPrincipalInvestigator,

In