A.I.EVOLUTIONDATATRENDSINHEALTHCARE’SCURRENTMARKET
JohnFrownfelter,MD,FACPCMIO,Jvion
“ToKnow,orNottoKnow…”
• Demandforanalyticsisatanall-timehigh:
• “HowamIperformingagainstqualitystandards?”
• “Howdoweidentifythepatientswhoaregoingtodieinthenext30days?”
• “Howdoweidentifythepatientswithincreasingriskforadmissiontothehospital?”
• “Howdoweknowourinterventionsareworking?”
TheUrgencyforKnowing• Valuebasedpurchasing/ACOContracts
• HowcanImitigatetherisk?
• PopulationHealthManagement• Assessingandinterveningonrisingrisk
• CMSPenalties• Readmissions• Qualitymeasuresof“never”eventslikeCAUTI,VTE
• MACRA
Maslow’sHierarchyofData
Learn/Optimize
Aggregate/ Label
Explore/Transform
Move/Store
Collect
AI, Deep
LearningA/B Testing,
Experimentation, Simple ML Algorithms
Analytics, Metrics, Segments, Aggregates, Features, Training
Data
Cleaning, Anomaly Detection, Prep
Reliable Data Flow, Infrastructure, Pipelines, ETL, Structured and Unstructured Data Storage
Instrumental, Logging, Sensors, External Data, User Generated Content
WhatisAI:AWorkingDefinition
• ArtificialIntelligence(AI):computersperformtasksthatareusuallyassumedtorequirehumanintelligence
• Accenture:ArtificialIntelligence(AI):healthcare'snewnervoussystem
• “AIinhealthrepresentsacollectionofmultipletechnologiesenablingmachinestosense,comprehend,actandlearn sotheycanperformadministrativeandclinicalhealthcarefunctions.Unlikelegacytechnologiesthatareonlyalgorithms/toolsthatcomplementahuman,healthAItodaycantrulyaugmenthumanactivity.”
• AnAImachinecanacceptinformationaboutaproblemfromitssurroundings,generateinsightsbasedonthisdata,anddeterminethebestcourseofactionthatwillleadtoadesiredoutcome
TheApplicationofAIwithinHealthcare:Top10
Robot-AssistedSurgery$40B
VirtualNursingAssistants$20B
AdministrativeWorkflow
Assistance$18B
FraudDetection$17B
DosageErrorReduction$16B
ConnectedMachines$14B
ClinicalTrialParticipant
Identifier$13B
PreliminaryDiagnosis$5B
AutomatedImageDiagnosis
$3B
Cybersecurity$2B
PerceptionsandChallenges
• June2018:TheAmericanMedicalAssociationpasseditsfirstpolicyonso-called"augmentedintelligence,"encouragingthedevelopmentofaugmentedintelligencetoolsthatarefreeofbiasandimprovepatientoutcomesandphysiciansatisfaction.
• RobertPearlwrites“thebiggestbarriertoartificialintelligenceinmedicineisn’tmathematics.Rather,it’samedicalculture….”
PilotOutcomes—NorthwestMedicalSpecialtiesAppliedAIforOncology
Sibel Blau,MD,President/CEO,QualityCancerCareAllianceMedicalDirector,OncologyDivisionNorthwestMedicalSpecialties
AmyEllis,Director,QualityandValueBasedCareNorthwestMedicalSpecialties,PLLC
CompositionoftheOncologySpecialtyVectorsONCOLOGY VECTORS DEEP DIVE
Vector Description
30 Day Mortality Patients at risk of mortality within 30 days of prediction
30 Day Pain Management Patients at risk of having severe/moderate pain within 30 days
6 Month Depression Patients at risk of having a depression diagnosis within 6 months
6 Month Deterioration Patients at risk of deterioration of ADL levels (at least 2 levels) within 6 months
30 Day Avoidable Admission Patients at risk of an avoidable IP admission within 30 days
30 Day ED Visit Patients at risk of an ED visit within 30 days
Readmission Patients at multiple admissions within 3 months
OncologyVectorsOverview
Jvion©2018Confidential
OncologyPracticePatient
EigenUniverse
DataTransposition
HL7
AIProcessing
DailyPropensities,RiskFactors,
Recommendations
MachineOutput
HL7,Extracts
EMRIntegratio
n
ClinicalWorkflow
HL7,Extracts
DataTranspositio
n
Oncology Vectors: Operational to Direction of Impact
• Up to 30% reductionin loss of function/ADLs (ECOG)
6-month Deterioration
• 22% increase in depression diagnoses
6-month Depression
• 33% reduction in moderate and severe pain
30-day Pain Management
StartwiththeWHY…TheRealImpact
CASE STUDY
Vector:Oncology30DayPainManagement
3300 3400 3500 3600 3700
# of patients reporting severe to moderate pain
Baseline Post-Jvion
20406080
Average Percentage of Patients Reporting Severe Pain, High &
Medium Risk Groups
Post-Jvion
Pre-Jvion
Ø 184 patients experienced improved pain management on average per month
Ø 552 total patients impacted post-JVION
0
1000
2000
3000
Reduction in Patients Reporting Severe Pain, Post-JVION
highrisk
mediumrisk
0
100
200
300
400
500
600
700
800
Jul-17 Aug-17 Sep-17 Oct-17 Nov-17 Dec-17 Jan-18 Feb-18 Mar-18 Apr-18 May-18 Jun-18 Jul-18 Aug-18 Sep-18
CountofPatientswithSeverePainperMonthatNWMS
709.7
664.5635.7
MortalityMetricsHospice&PalliativeCareReferrals
0.1
0.5
0.00.10.10.20.20.30.30.40.40.50.5
PreJvion PostJvion
HospiceReferralsper1,000patientspermonth
8.4
11.3
0.0
2.0
4.0
6.0
8.0
10.0
12.0
PreJvion PostJvion
PalliativeCareReferralsandSupportiveCareConsultsper1,000patientspermonth
225.0% increase in rate 35.3% increase in rate
0
50
100
150
200
250
300
350
400
450
500
Jul-17 Aug-17 Sep-17 Oct-17 Nov-17 Dec-17 Jan-18 Feb-18 Mar-18 Apr-18 May-18 Jun-18 Jul-18 Aug-18 Sep-18
CountofPatientswithDepressionperMonthatNWMS
399.0
374.7325.7
PilotOutcomes—TheCenterforCancerandBloodDisorders
AppliedAIforOncology
RayPage,DO,PhD,President&DirectorofResearchTheCenterforCancerandBloodDisorders
Oncology Vectors: Operational to Direction of Impact
• 17% reduction in loss of function/ADLs (ECOG)
6-month Deterioration
• 33% increase in depression diagnoses
6-month Depression
• 28% reduction in moderate and severe pain
30-day Pain Management
StartwiththeWHY…TheRealImpact
CASE STUDY - CCBD
Vector:Oncology30DayPainManagement
0 200 400 600 800
# of patients reporting severe to
moderate pain
Baseline Post-Jvion
20
30
40
Average Percentage of Patients Reporting Severe Pain, High Risk
Group
Post-Jvion
Pre-Jvion
Ø 71 patients experienced improved pain management on average per month
Ø 499 total patients impacted post-JVION
0
200
400
600
Reduction in Patients Reporting Severe Pain, Post-JVION
0
100
200
300
400
500
600
700
800
Jul-17 Aug-17 Sep-17 Oct-17 Nov-17 Dec-17 Jan-18 Feb-18 Mar-18 Apr-18 May-18 Jun-18 Jul-18 Aug-18 Sep-18
CountofPatientswithSeverePainperMonthatTCCB
635.0 613.5
510.3
MortalityMetricsAveragesper1,000patientsperMonth
0.01
0.03
0.00
0.01
0.01
0.02
0.02
0.03
0.03
PreJvion PostJvion
HospiceReferralsper1,000patientspermonth
0.03
0.08
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
PreJvion PostJvion
PalliativeCareReferralsper1,000patientsperMonth
113.3% increase in rate 218.8% increase in rate
0
50
100
150
200
250
Jul-17 Aug-17 Sep-17 Oct-17 Nov-17 Dec-17 Jan-18 Feb-18 Mar-18 Apr-18 May-18 Jun-18 Jul-18 Aug-18 Sep-18
CountofPatientswithDepressionperMonthatTCCB
85.0 94.8 excludingDec.2017 88.3
CCBD– 30dayMortality
32
CCBD– 30dayDepression
33