thurs 755am pacificb general artificialintel blau ellis ...€¦ · •“ai in health represents a...

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