Ward slides for AMGA ACO Collaborative - 2011-07-14 final ......Jul 14, 2011 · Richard E. Ward, MD, MBA Reward Health Sciences, Inc. July 14, 2011 Health Sciences ... blocks in

Developing the Economic Model for a Successful ACO

AMGA ACO Learning CollaborativeSwissotel, Chicago

Richard E. Ward, MD, MBAReward Health Sciences, Inc.

July 14, 2011

Health SciencesHealth SciencesREWARDREWARD

Copyrighted 2011, Reward Health Sciences, Inc.1

Outline

• How can ACOs reduce cost• ACO structure dilemmas• IT investment priorities• Using analytic models

– Population Management – Provider Incentives– ACO Financial Models


The ACO Stack


Clinician Workstation‐ Results‐ Profiles‐ To Do List‐ Guidelines

Relative Strength of Sources of Cost Savings for ACOs (illustrative)

Copyrighted 2011, Reward Health Sciences, Inc.

Cost Impact

Reduce Use of Low Value Services of Specialists and

Facilities

PCP Referral Influence

Reduce Rate of Avoidable Clinical

Events

Patient Self‐Management Support

Care Coordination

Reduce Resources Per Clinical ServiceLean

Reduce Duplication of Services

Clinical Decision Support

Health Information Exchange

Provider Consolidation increasing Market Power

Increase Price per Clinical Service or

Episode

Delivery System Transformation

Patient‐Centered Medical Home

and

Accountable Care Organization

and

Meaningful Use of Health Information

Technology

4

Fundamental Structural Dilemma #1


HealthPlan A

HealthPlan B

HealthPlan C

Provider 1

Patients

Provider 2

Patients

Provider 3

Patients

Free Rider Problem• “If Plan A invests in core

process improvement and HIT for its providers, the other plans will receive the savings without bearing the cost. So they will gain advantage.”

Scale Problem• “If Plan A puts it’s own

care managers into clinics of its providers to serve only members of Plan A, there is not enough work to keep the care manager busy.”

Externality Problem• “If Provider 1 invests its

own resources in process improvement and IT, the savings accrue to the health plans.”

Many‐to‐Many Relationshipsbetween Plans and Providers

5

• Medicare• Meaningful Use• ACO Gain Sharing• Comparative Effectiveness• Demonstrations

• Some non‐profit Blues plans with high local market share

• PCMH• P4P• Organized Systems of Care

• Health Systems• Kaiser Permanente,

Geisinger• Plan‐initiated

• HealthSpring?• Disease‐specific

• McKesson/US Oncology?• Coops formed with reform bill

funding?

2 Ways Out of Structural Dilemma #1


Dominant Payer with Resources and a Social Mission

(Willing to invest even if some benefits accrue to other payers)

Provider with Mostly‐Exclusive Relationship with Payer

(May start as a niche but then grow by outcompeting others)

6

Patients that have aCare Relationship

with PCMH

Patient CenteredMedical Home

SpecialtyGroup Practice

Hospital & Specialists

HomeHealth

Care MgmtVendor

NursingHome

DMESupplier

EmployeeHealth Clinic

Hospital

Specialist

Specialist



7



Every party’s cost savingsis another party’s revenue loss...

…and they are not going to be happy.

8

Pre‐1990’s Delivery System Model


Primary Care

Specialist HospitalHealth Plan

9

Idealized 1990’s “Health System” or “Staff‐Model HMO” Model


Primary Care

Specialist HospitalHealth Plan

Health System

10

Conventional 2000’s Organizational Alignment


Primary Care

Specialist Hospital

Hospitalist

Health Plan

Care Manager

PhysicianOrganization

HospitalSystem

Payer

11

Integrated Delivery System ACO Model


Primary Care

Specialist Hospital

Hospitalist

Health Plan

Care Manager

Accountable Care Organization

12

Primary Care‐based ACO Model


Primary Care

Specialist Hospital

Hospitalist

Health Plan

Care Manager

Primary Care‐based ACO

Co‐Managed Service Lines

13

Alignment of Specialists Depends on the Focus of their Practice


General Internists

Cardiovascular Surgeons

Nurse Care Managers

Cardiologists focused on Heart Failure

Inpatient Care andProcedures

Mid‐Levels focused on ambulatory care

Interventional Cardiologists

Radiologists

Pathologists

Anesthesiologistsin OR

Emergency

Endocrinologistsfocused on Diabetes

Physiatrists focused on surgical rehab

Pulmonologist focused on COPD

Rheumatologists focused on chronic

Osteoarthritis

Pulmonologist in Critical Care

OrthopedicSurgeonsSurgical

Oncologist

Radiation Oncologist

Medical Oncologist

Physiatrists focused on chronic back pain

Cardiologists focused on CCU

Heart & Vascular Service Line Co‐management LLP

Cancer Service LineCo‐management LLP

Psychiatrists focused on depression

Anesthesiologistsfocused on pain mgmt

Ambulatory Care and Population Management

Family Practitioners

Pediatricians

Hospitalists

Transitions of Care &Resource Stewardship

Length of Stay &Referral “Keepage”

14

“Cooperative” Model


Primary Care

Specialist Hospital

Hospitalist

Health Plan

Cooperative

Care Manager

Look familiar?

15

Health Information Technology


Accountable CarePatient CenteredPopulationProcess

Guidelines & ProtocolsMeasures

Going PaperlessClinical Data Accessibility, Efficiency, SecurityOld Vision

New Vision

16

Health Information Technology


Process

DataOld Vision

New Vision

17

Benchmarks

Goals

Quality & Cost

PerformanceAnalysisLiterature

Expert Opinion

BestPractices

Data

Outcomes

Process

Feedback

IncentivesProtocols

Guide

lines Implementation

HealthCare

Care‐Delivery

Care‐Planning


Systems to Enable Process Transformation

HealthCare

Care‐Delivery

Care‐Planning


Leverage Workflow Automation / Business Process Mgmt Technology used in other industries

TightlyIntegrated

Care Planning ToolsPatient CenteredProblem Oriented

SmartPopulation

Care ProcessManagement Tools

Physician controlledMeasurableCoordination

19

Unstructured

• Free text• Dictated and Transcribed• Dictated and voice‐

recognized• Document Images• Optical Character

Recognition

• Drawings• Clinical Images• Sounds

• Human readable

Passively Structured

• Text‐to‐code logic• Commands to include text blocks in notes

• Loosely structured messages

• Human readable with more consistent formatting

• Case finding

Actively Structured

• Registry• Questionnaire• Form‐based Template Charting

• Problem‐oriented clinical documentation templates

• Tightly structured messages

• Human readable with most consistent formatting

• Reminders and alerts• Performance measures• Comparative effectiveness

Health Information


Clinical DataRepository

Analytical Data Repository

Admin SystemsAdmin Systems Clinical Ancillary SystemsClinical Ancillary Systems

Clinician Workstation‐ Results‐ Profiles‐ To Do List‐ Guidelines

IT Framework to Support ACOs

Analysis & Reporting

• Quality• Episode Profiling• Provider Network Analytics• Registry• Research

Clinical Workstation• Results• Profiles• In Box• Schedule• Guidelines

Clinical Process Mgmt

• Process Designer• Process Simulation• Process Monitoring• Questionnaire Designer• Order/Result Mgmt• Clinical Protocol Mgmt• Care Relationship Mgmt• Call Center Integration• E‐mail, Text & IM

Patient Apps• Results• Coaching• Telemedicine• Care Plan• Questionnaire

Care Planning• Smart Templates• Orders• Clinical Documentation

*US patents #7020618, 7707057


Analytic Data Repository

RawVersioned

Data

Source Systems

Reports &Reporting

Applications

out

Analytic Data Repository Framework to Support ACOs

Scheduling

Admit, Discharge,Transfer (ADT)

Billing

MedicationAdministration

Operating Room

Credentialing

Etc.

in

Data Derivation Engines & Services

Disease ID Risk ScoresGaps in Care Episodes of Care

Clinical Data Repository

Cubes & Other Summary

Data Structures

Care Relationships

Specialty / Peers

ReferralRelationships Etc.

Derived data

Analyzable Data

• Normalized• Documented• With derived

entities and attributes



MODELSREPORTS vs.

Looking back Looking ahead

23

ChronicConditionsWellness

Concerns& Symptoms

Acute Conditions

ElectiveSurgical Conditions

Complex Catastrophic Conditions

Continuum of Patient Needs

Using Models for Care Management


25

Is Care Management Effective?• Are drugs effective?• Is a scalpel effective?

• Which population?• What point in time?• What intervention?• What outcomes of interest?• What time horizon?• What evidence threshold?

It depends


26

TARGETEDHOLISTICCompeting Intervention Design Philosophies

• Easier to design• Respects professionalism• Addresses patient complexity• Difficult to evaluate

Many “triggers”

General Assessment

Multi‐Issue Care Plan

Intervention Periodas Coach Evolves Goalsand Revised Care Plan

• Consistent intervention process enables process improvement

• Targeting protocol can be applied to comparison population for evaluation

Targeting of PatientsBased on Objective CriteriaBased on Opportunity to

Benefit from aparticular intervention

Outreach Protocol

Intervention Protocol


27

Using Intervention Models to Explore Alternative Interventions

Care Transition Nurse On Site Telephonic

Identified Population/Spend $100 $100

Patients Identified in when still in hospital

$100 $48

Target Rate$100 $41

Reach and Engagement Rate

Effectiveness Rate in avoiding need for readmission

$65 $13

Total Gross Savings $20 $2

100% 48%

100% 86%

65% 32%

30% 15%

IllustrativeCopyrighted 2011, Reward Health Sciences, Inc.

28

Intervention Design

Cause‐Effect Model

includes

Process Model

includes

Intervention Model

informsinformsEvaluation

Plan informs

BusinessProcessWorkflow

Diagram (BPD)

informs

Clinical ProgramOperations

orchestrates

Activity Datacreates

enablesextrapolation of

Calculated ActualOutcomes

toProjected Outcomes

For AlternativeIntervention Designs

enablescalculation of

supports assumptions of

confirms plausibility of

Effect Measurement

informs

informs


29

Illustrative


30

Number of IP admissions per 1000 members identified with CHF, by percentile of risk score

0

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

4,500

5,000

0102030405060708090100

Percentile of Symmetry risk score

IP A

dmit

Rat

e pe

r 100

0

Predicted rate per 1000

Overall IP Rate

Illustrative


31

Diabetes Disease Management

(1,000,000)

-

1,000,000

2,000,000

3,000,000

4,000,000

5,000,000

0% 5% 10%

15%

20%

25%

30%

35%

40%

45%

50%

55%

60%

65%

70%

75%

80%

85%

90%

95%

100%

Finding Target Penetration that Yields Max Net Savings:Maximizing Beneficial Impact for Members for the Amount Spent

Gross Savings

Cost

Net Savings

Dollars

41%

Fixed Cost

Illustrative

Target Penetration Rate (as % of Diabetes population)


32

Chronic Disease Management

(0.15)

(0.10)

(0.05)

-

0.05

0.10

0.15

0.20

0.25

- 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45

Variable Cost PMPM

Net

Sav

ings

PM

PM IHDCHFDiabetesCOPDAsthma

47% of Ischemic Heart Disease

87% of CongestiveHeart Failure

41% of Diabetes

34% ofCOPD 20% of Asthma

Max Net Savings Signature Illustrative


33

Dynamic Models

• Thinking like an accountant analyzing accounts receivable


34

-$5M

$0M

$5M

$10M

$15M

$20M

Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4

Benefit Cost SavingsOperational CostsInvestment CostsQuarterly Economic Impact

Dynamic Models

Quarterly Economic Impact

2009 2010 2011 2012 2013 2014

Case Management

Illustrative


35

Dynamic Models


2009 2010 2011 2012 2013 2014

ILLUSTRATION

Chronic Condition Management

‐$1.0M

‐$0.5M

$0.0M

$0.5M

$1.0M

$1.5M

$2.0M

$2.5M

$3.0M

$3.5M

Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4

Benefit Cost Savings

Operational Costs

Investment Costs



36

Analyzing UncertaintyUsing Monte Carlo Simulation

Assumptions

Calculations

90% Interval of Uncertainty


37

Chronic Condition Management—Sensitivity Analysis

2014 Cumulative Net Savings Frequency Distribution 2014 Cumulative Net Savings Variable Sensitivity

Contribution to Variance

Illustrative

1%

1%

1%

1%

2%

17%

71%

7%

0% 20% 40% 60% 80%

Other

Average Length of Regular

Engagement Phone Calls (min)

MA PPO Annual Inflation (Program

Costs) Growth Rate

MA PPO Annual Medical Spend

Growth Rate Above Inflation

Double Counting Assumption

Engagement Rate (% of reached

members engaged)

Member Reach Rate (% of targeted

members reached)

Total Spend Reduction for Engaged

Members

0

200

400

600

800

1,000

1,200

‐$10M $4M $18M $32M $46M $60M

Freq

uency


38

Example of “Hurricane Diagram” WCM Solution Cumulative Net Savings

90%Confidence

Note: Based on a Monte Carlo analysis with 10,000 trials, and triangular distributions on 72 input variables for entire portfolio

Range of Outcomes—Cumulative Portfolio Net Savings

ILLUSTRATION

‐$20M

$M

$20M

$40M

$60M

$80M

$100M

$120M

$140M

Jun‐10 Dec‐10 Jun‐11 Dec‐11 Jun‐12 Dec‐12 Jun‐13 Dec‐13 Jun‐14 Dec‐14


39

InghamInghamInghamInghamInghamInghamInghamInghamIngham

KalamazooKalamazooKalamazooKalamazooKalamazooKalamazooKalamazooKalamazooKalamazoo

KentKentKentKentKentKentKentKentKent

OaklandOaklandOaklandOaklandOaklandOaklandOaklandOaklandOakland

WashtenawWashtenawWashtenawWashtenawWashtenawWashtenawWashtenawWashtenawWashtenaw WayneWayneWayneWayneWayneWayneWayneWayneWayne

Modeling Geographically‐Sensitive Interventions

County # Facilities # NCMs1 Annual Net

SavingsEngaged LocallyOakland County 85 10 $ xWayne County 91 8 $ xKent County 22 4 $ xWashtenaw

County 19 3 $ x

Ingham County 7 2 $ xKalamazoo

County 12 2 $ x

Engaged TelephonicallyAll Other

Counties 364 $ x

= Counties targeted locally

1 NCMs = Nurse Care Managers

ILLUSTRATIVE

In‐HospitalCare Transition Nurse


40

Projected benefit cost savingsAnnual savings by initiative category

$5,006

$6,281

$7,530

$8,472

$0

$1,000

$2,000

$3,000

$4,000

$5,000

$6,000

$7,000

$8,000

$9,000

2010 2011 2012 2013

Ben

efit

cost

sav

ings

($k)

Service utilization ConditionClinical IT Core clinical processNew group Planned

Projected benefit cost savingsAnnual savings by initiative category as % of total benefit cost

0.23% 0.24% 0.24% 0.24%

0.18%

0.23% 0.25% 0.25%

0.25% 0.25% 0.25% 0.25%

0.35%

0.40%0.43% 0.44%

0.00% 0.00% 0.00% 0.00%0.0%

0.1%

0.1%

0.2%

0.2%

0.3%

0.3%

0.4%

0.4%

0.5%

0.5%

2010 2011 2012 2013

Ben

efit

cost

sav

ings

(%)

Service utilization ConditionClinical IT Core clinical processNew group Planned

Modeling Provider Incentive ProgramsSavings for Customer X for 41 Initiatives in the BCBSM Physician Group Incentive Program

ILLUSTRATIVECopyrighted 2011, Reward Health Sciences, Inc.


MODELSREPORTS vs.

Looking back Looking ahead

41

42

• Comparison group is not truly comparable

• Noise > Signal

• Noise = “common cause” or “random” variation in people and their response to disease and treatment

BIASVARIATION

The Two Key Challenges to Measurement


43

6,533

3,450

-1,0002,0003,0004,0005,0006,0007,0008,0009,000

10,000

Pre Intervention (3 months) Post Intervention (3 month)

Aver

age

Cost

Per

Cas

e (P

MPM

)

Regression to the Mean

47.1%Reduction!

$3,083SavingsPer Case!

n=11,768

Case Management in Senior PopulationCost per Case before and after referral

Illustrative


44*Post date ranges in relation to 5‐days after targeting.

-

2,000

4,000

6,000

8,000

10,000

12,000

14,000

16,000

Pre61-90

Pre31-60

Pre0-30

Post'0-30

Post31-60

Post61-90

Post91-120

Post121-150

Post151-180

Post181-210

Days in Relation to Targeting for Case Management*

Cos

t Per

Mem

ber P

er M

onth

Engaged

Illustrative

n=11,768




45*Post date ranges in relation to 5‐days after targeting.

Engaged

Not Engaged-

2,000

4,000

6,000

8,000

10,000

12,000

14,000

16,000

Pre61-90

Pre31-60

Pre0-30

Post'0-30

Post31-60

Post61-90

Post91-120

Post121-150

Post151-180

Post181-210

Days in Relation to Targeting for Case Management*

Cos

t Per

Mem

ber P

er M

onth

Illustrative

n=11,768




46

0

50

100

150

200

250

300

350M

on 1

Mon

2

Mon

3

Mon

4

Mon

5

Mon

6

Mon

7

Mon

8

Mon

9

Mon

10

Mon

11

Mon

12

Mon

13

Mon

14

Mon

15

Mon

16

Mon

17

Mon

18

Mon

19

Mon

20

Mon

21

Mon

22

Mon

23

Mon

24

Mon

25

Mon

26

Mon

27

Mon

28

Pre-Intervention Actual

Pre-Intervention Trend

Expected Post-Intervention Trend

Post-Intevention Actual

Solution = Outcomes Monitoring with “Re‐qualification”

Ramp‐Up Intervention Steady State

IllustrativeRegression to the Mean


47

0

50

100

150

200

250

300

350M

on 1

Mon

2

Mon

3

Mon

4

Mon

5

Mon

6

Mon

7

Mon

8

Mon

9

Mon

10

Mon

11

Mon

12

Mon

13

Mon

14

Mon

15

Mon

16

Mon

17

Mon

18

Mon

19

Mon

20

Mon

21

Mon

22

Mon

23

Mon

24

Mon

25

Mon

26

Mon

27

Mon

28

Pre-Intervention Actual

Pre-Intervention Trend

Expected Post-Intervention Trend

Post-Intevention Actual

Solution = Outcomes Monitoring with “Re‐qualification”

Ramp‐Up Intervention Steady State

IllustrativeRegression to the Mean


48

Applying Outcomes Monitoring to aVendor‐delivered Disease Mgmt Program


49

Using Statistical Models


Dynamic ACO Financial Model

• ACO gets into financial trouble if their utilization efficiency success outpaces the market conversion to performance‐based reimbursement and the ACO’s efforts to reduce its fixed cost base.

• Deals with health plans can be structured to reduce or share this transition risk.

• The key is to create a dynamic model of the economics from all parties’ perspectives, with believable assumptions and the right balance of simplicity vs. detail.


ACO Financial ModelAccountable Care Organization

Out of Pocket (copay $, coins %)

Prem

ium sh

are

Premium

P4P

FFS

P4P

Sal

Cap

Cap

P4P

Cap

Cap

Employer

Person

ACO Legal / Contracting Entity

Bon

Sal

Bon

HospitalPCP

HospitalChain

SpecialistIncentive

PhysicianOrganization

PrimaryPractices

SpecialtyPractices

HealthPlan

HSA & Sal


Summary

• ACO success requires attention to “ACO Stack”• Primary Care vs. Health System Model ACO structure• IT emphasis on care planning & care delivery coordination• Importance of actively structured data• Importance of models, not just reports• Models should address uncertainty and dynamics


Thank You!


Questions

Contact Info:Richard E. Ward, MD, [email protected]

519‐817‐8300

53

Documents

Ward slides for AMGA ACO Collaborative - 2011-07-14 final ......Jul 14, 2011 · Richard E. Ward, MD, MBA Reward Health Sciences, Inc. July 14, 2011 Health Sciences ... blocks in