driving decisions from predictive modeling
DESCRIPTION
TRANSCRIPT
Driving Decisions from
Predictive Modeling
O r l a n d o , F l o r i d a
DATE : January 27, 2010
2
Speaker Bios
Introduction
Anand Rao is a Partner at Diamond Management & Technology Consultants. Anand has more than twenty
years of experience in using advanced techniques, such as predictive modeling, agent-based simulation,
and system dynamic techniques to analyze decision making situations. He has advised clients on a
number of different aspects of customer experience and value management; behavioral economics and
interventions; large scale transformation strategy, design, and execution. Anand is the lead proponent of
the Belief-Desire-Intention agent modeling paradigm and was recognized for his contribution in this field
with the Most Influential Paper Award for the decade by AAMAS in 2007. He has co-authored a number of
papers, has written four books on building intelligent systems and is a frequent speaker at conferences on
intelligent systems, behavioral economics, and advanced analytical techniques
([email protected] and www.anand-rao.com)
Richard Findlay is a Practice Director in Diamond‟s Healthcare practice with over 25 years of experience
across the healthcare industry. He has focused on developing for clients business strategies that create
competitive advantage through the optimal use of Information Technology. His portfolio of expertise spans
the value chain of operations with a particular focus on informatics, sales and marketing, supply chain,
and clinical development. Within the industry Richard held senior executive positions with SmithKline
Beecham and Abbott Laboratories. He is recognized as a leading authority on the future of healthcare and
informatics frequently publishing and speaking at conferences on leading edge developments
Amaresh has helped Fortune 500 companies in multiple industries to use bottoms-up data analytics in
strategic decision making. His work has focused on developing growth strategies, defining market entry
plans, understanding customer behavior to increase profitability, improve marketing efficiency, developing
operations strategy and streamlining distribution. Amaresh founded Diamond‟s Information and Analytics
practice and set up its delivery center in Mumbai, which he helps to manage. He is the editor of Diamond‟s
information analytics blog (www.diamondinfoanalytics.com) and has written white papers on customer
service, marketing segmentation and behavioral economics,. Amaresh holds a Masters degree in
Transportation Systems Engineering from the University of Texas at Austin
3
Content
Market Context
Our Predictive Modeling Thesis
Strategic Applications: System Dynamics Modeling &
Demonstration
Operational Applications: DRIVE
Behavioral Economics in Healthcare
Questions
Content
4
Government• Healthcare reform
• Focus on interoperability,
CPOE, and EHRs/ EMRs
• Fiscal stimulus
Rising Healthcare Costs
• Employer costs
• Employee costs
• Sicker population
– Aging & Young
• New technologyTechnology
• Data collection &
sharing
• Informatics &
analysis
• Presentation &
touchpoints
• Decision support
& planning
Healthcare Market
Market Forces
• Convergence of
value chain
• New business
models
• Changing
industry
structureEconomic Recession
• Employer bankruptcies
and cost reduction
• Payers losing group
business
• Reduced demand for
providers
• Consolidation
Consumerism
• Consumer Directed
Health Plans
• Transparency
• Fear of Change
Market Context
Disruptive forces shaping Healthcare market
5
• Stock market slide has hurt payer
reserves
• Projected EPS growth ~5%, well
below 2004-07 gains (S&P)
• Decline in group business
Payers
• Growing numbers of unemployed
and projected uninsured
• Unfunded retirement burden
(Medicare and Social Security)
Consumers
• Falloff in non-essential procedures
• Rising bad debt
• Focus on HMO operating model
Providers
• Carry most of healthcare cost
burden on top of other economic
challenges
• Pressure to reduce costs
Employers
• Rx rates down
• Looking to consolidate for scale
• Restructuring to address new go to
market strategies
Suppliers
Market Context
Impact of Economic Recession
6
MMA
creates
HSAs &
mandates
eRecords
HIPAA
establishes
standards –
compliance
by 2005
Executive Order
mandates
requirements
for technology,
transparency
and incentives
Tax-Free
Healthcare
Savings, Access,
and Portability
Act increases
financial
attractiveness of
HSAs
Universal Health
Care Choice and
Access Act to allow
use of pre-tax dollars
for individual health
premiums
HSA Improvement
and Expansion Act
increases limits
and flexibility in
use and funding of
HSAs
Tax Relief and
Health Care
Act increases
flexibility and
limits for
funding HSAs
NIST will foster
the
development of
a national
infrastructure
to share health
data
Source: Library of Congress – Thomas search.
2003 Healthcare Legislation Timeline 2010
Market Context
Increasing Government Activism
Stimulus and
Budget
Packages.
HIT spend $20B.
Pharma Pricing
restraints
Protech Act
in
Committee
EMEDS
Act
7
Reform is emerging in 2010.
Major alignments will result in:a. Greater percentage of individual plans vs. group membership.
b. Greater interaction with government data bases and programs for all value
chain constituents.
Predictive modeling will need to assessa. Legislation impact on member choices
b. Legislation choices for non member options
c. Impact and inference of “High Risk” pool
d. Payers new product opportunities
e. Payers need to re-segment the market place
f. Specifics around impacts of closing the “doughnut hole” for payers &
Pharma
Market Context
Greatest force for change in next 5 years is Healthcare Reform
8
„Holistic‟ Advice
& Financial
Planning
Investment
Advice
Integrated
Individual
Risk Mgmt
Integrated
Portfolio &
Benefits
Mgmt
Hold Funds
and Manage
Investments
Process
TransactionsPlatform
Alternative
Risk/ Capital
Markets
Distributor Risk AggregatorAdviser ManufacturerBack-officeAdministrator
RiskTransferor
Market Context
Reform will further impact business models to change
9
Clinical Information Chain
Collection, storage,
aggregation and
sharing from and to
multiple sources
Distilling large data
sets to guide
decisions for care and
business operations
Delivering the
information back to
providers and patients
conveniently and
coherently
Driving behavioral
change to improve
health outcomes
Informatics
&
Analytics
Presentation
&
Touchpoints
Decision
Support &
Planning
Data
Collection
Market Context
Different types of value are added at each step in the clinical information chain
10
Content
Market Context
Our Predictive Modeling Thesis
Strategic Applications: System Dynamics Modeling &
Demonstration
Operational Applications: DRIVE
Behavioral Economics in Healthcare
Questions
Content
11
Strategic and Operational predictive modeling need different
tools and analysis approaches
Integration of multiple data sources, especially third party
data, provides better predictions
Statistical techniques are mature and normally not worth the
incremental investment dollar
Good data visualization leads to smarter decisions
Delivering the prediction at the point of decision making is
critical
Architecture is critical
Prototype, Pilot, Scale
Our Predictive Modeling Thesis
Diamond‟s Predictive Modeling Thesis
12
Analytical Techniques Simulation TechniquesPrediction e.g., Linear & Logistic
Regression
Segmentation e.g., CHAID & Factor
Analysis
Optimization e.g., genetic algorithms
and linear programming
Discrete-event Simulation
Agent Based
System Dynamics
Dynamic Systems
Our Predictive Modeling Thesis
Strategic Vs. Operational Predictive Modeling Tools
Deterministic
Predict equilibrium point
Linear flows
Point solutions
E.g., claims fraud, segmentation &
targeting, efficacy of disease
management program
Learn drivers of
stock & flow over time
Feedback loops
Systemic understanding
E.g., Disease epidemiology, patient
flow models, impact of public policy
reform
Operational Decisions Strategic Decisions
13
Integrate Multiple Data Sources
Which data would you look for to predict dentist potential?
Our Predictive Modeling Thesis
14
1880s: Linear Regression proposed by Galton
1944: Logistic Regression proposed by Berkson
1954: Systems Dynamics developed by Forrester
1969: Backpropogration method in neural networks
1976: SAS Founded by Jim Goodnight
1993: R Open source statistical environment launched
Our Predictive Modeling Thesis
Mature Statistical Techniques & Tools
15Our Predictive Modeling Thesis
Good data visualization leads to smarter decisions
Dr. John Snow‟s Visualization at 40 Broad Street (1854)
Convinced city officials that cholera is a water borne disease
16Our Predictive Modeling Thesis
40 Broad Street Water Pump
Photo Credit: Miles Dowsett (www.milesdowsett.com)
17Our Predictive Modeling Thesis
Delivering predictions at the point of decision making
Forecast
Logic
Application
Handheld
Data checks Data Servers
Inventory/
Hole Count
Current Price
Area Sales Managers
Place Order Next Day Delivery
ASM can Override if he feels necessary
Start Over
two days
later
FORECAST
Order Forecasting at Grocery Store
Reduction in OOS from 14% to 4%
18Our Predictive Modeling Thesis
Delivering predictions at the point of decision making
1. Select drug and connect to Payer to
determine eligibility
Rx
RxHub Direct
Connections
Prescriber
4. Send Rx to patient‟s pharmacy of
choice
Source: Archives of Internal Medicine Dec 2008.
E-Prescription
3.3% Increase in prescription of generic drugs when using an e-prescription system with formulary decision support.
Rx
Prescriber
2. Formulary/History brought to
provider
5. Renewal sent back to provider
Rx
Health Plans
Pharmacy
RxHub
3. Once Rx written,
drug interactions are
checked
19Our Predictive Modeling Thesis
Architecture is critical
Data
Aggregation
Engine
Analytical
Engine
Visualization & Reporting
Engine
Internal Data
External Data
Syndicated Data
Survey data
Predictive
Modeling
Clustering &
Segmentation
Optimization
Mapping
Graphing
Lists & Scores
Pivots
Access
Oracle
MS SQL
SAS
SPSS
R
Tableau
MS Excel
Mappoint
Business Objects/Cognos
To
ols
Bu
ild
ing
Blo
ck
s
CR
M, S
FA
, M
obile
Devic
e Inte
gra
tion
Source: Archives of Internal Medicine Dec 2008.
20Our Predictive Modeling Thesis
Prototype, Pilot, Scale
Prototype Pilot Scale
Choose pilot area
Pilot measurement
framework
Train and launch pilot
Gather feedback on
rollout process
Update tactical
elements based pilot
learnings
Program integration
points for scaling the
prototype
Ongoing
measurement plan
Tasks
Define problem and
hypotheses
Identify datasets
Develop model and
output
Controlled pilot plan
Duration
2 months 3 months 4 months
Source: Archives of Internal Medicine Dec 2008.
21
Market Context
Our Predictive Modeling Thesis
Strategic Applications: System Dynamics Modeling &
Demonstration
Operational Applications: DRIVE
Behavioral Economics in Healthcare
Questions
Content
Content
22Strategic Applications
Strategic Vs. Operational Predictive Modeling Tools
Analytical TechniquesPrediction e.g., Linear & Logistic
Regression
Segmentation e.g., CHAID & Factor
Analysis
Optimization e.g., genetic algorithms
and linear programming
Deterministic
Predict equilibrium point
Linear flows
Point solutions
E.g., claims fraud, segmentation &
targeting, efficacy of disease
management program
Operational Decisions
Simulation TechniquesDiscrete-event Simulation
Agent Based
System Dynamics
Dynamic Systems
Learn drivers of
stock & flow over time
Feedback loops
Systemic understanding
E.g., Disease epidemiology, patient
flow models, impact of public policy
reform
Strategic Decisions
23
Source: From System Dynamics and Discrete Event to Practical Agent Based Modeling: Reasons, Techniques,
Tools by Borshchev, A., and Filippov, A.
Strategic Applications
Applicability of Simulation Techniques
Aggregates, Global Causal Dependencies, Feedback Dynamics,…
Individual objects, exact sized, distances, velocities, timings,…
“Discrete Event” (DE)
• Entities (passive objects)
• Flowcharts and/or transport networks
• Resources
Agent Based (AB)
• Active objects • Individual behavior
rules• Direct or indirect
interaction• Environment models
System Dynamics(SD)
• Levels (aggregates)• Stock-and-Flow diagrams• Feedback loops
Dynamic Dynamics(DS)
• Physical state variables• Block diagrams and/or
algebraic-differential equations
Mainly discrete Mainly Continuous
High AbstractionLess DetailsMacro LevelStrategic Level
Low AbstractionLess DetailsMicro LevelOperational Level
Middle AbstractionMedium DetailsMeso LevelTactical Level
24
Patient Flow Model
– Within a Provider across multiple departments
– Across primary, secondary, and community healthcare
Disease epidemiology
– Heart disease, Diabetes, HIV, cervical cancer, chlamydia infection
– Dengue fever and drug-resistant pneumococcal infections
Substance abuse epidemiology
– Heroin addition, cocaine prevalence and tobacco reduction policy
Healthcare capacity and delivery
Interactions between public health capacity and disease epidemiology
Strategic Applications
Application of Simulation Techniques in Healthcare
25Strategic Applications
Patient Flow Model – Example
26Strategic Applications
Patient Flow Model – Example
27Strategic Applications
Two Case Studies
Diabetes Management Healthcare Policy (UK)
28
0
1
2
3
4
5
6
7
8
9
1935 1965 1982 1999 2015 2022
Rem
ain
ing
Years
Independent
Disabled
53%
34%
66%
47%
38%
62%
72%
77%
35%
Strategic Applications
Compressed Morbidity: Longer life and fewer disabled years
Source: The Aging Boom: Demographic Trends and Policy Implications, Department of Elder Affairs State of Florida (H) Committee on
Healthy Seniors, January 22, 2008 Charlie Crist Governor, State of Florida Statistics
Life Expectancy at Age 85
29
Incidence of chronic disease increases with age, however;
Improvements in disease management have reduced the
disabling effects of morbidity.
Therefore, even as there are increases in chronic disease
there are reductions in disability at advanced ages;
Leading to longer independent life-spans.
Strategic Applications
Behind the Numbers: Compressed Morbidity
Source: The Aging Boom: Demographic Trends and Policy Implications, Department of Elder Affairs State of Florida (H) Committee on
Healthy Seniors, January 22, 2008 Charlie Crist Governor, State of Florida Statistics
30
Guiding Principles
Meaningful and Measurable
Simple and Easy to Use
Social Networking
Community Involvement
Interactive
Participation Incentives
Awareness Prevention Occurrence
Proposed Intervention
Self-care
Current Intervention
Primary-Care HospitalizationCritical
Care
Intensive
Care
Prototype Requirements
Multi-channel communication:
– Mobile via SMS
– PC (Personal Or Computer @ Tele-Centre)
6 months of execution for gathering data
Control group to evaluate efficiency & efficacy
Patient Healthcare Intervention Spectrum
Prototype Guiding Principles and Requirements
Strategic Applications
Where we saw opportunity
31Strategic Applications
The Cost of Diabetes: The Big Picture
32Strategic Applications
How we approached the opportunity
Ecosystem PartnersEcosystem Partners
Managed
Preventative
Medical CareUsing broadband as a
platform, managing patients
more effectively and
efficiently
• Better Citizen Health
• Improved Productivity
• Efficient and Effective
Healthcare Service
Delivery
Impact
• Better Citizen Health
• Improved Productivity
• Efficient and Effective
Healthcare Service
Delivery
Impact
• Diabetes is one of the most costly CDL
• The direct and indirect economic impacts to citizens and
governments can be very high
• 80% of diabetes can be well managed and easily controlled
Rationale
• Diabetes is one of the most costly CDL
• The direct and indirect economic impacts to citizens and
governments can be very high
• 80% of diabetes can be well managed and easily controlled
Rationale
• Patient Knowledge
• Capillary Blood Glucose Levels ( mmol/l)
• HbA1C %
• Hospitalization rate
• Body Mass Index
Key Performance Indicators
• Patient Knowledge
• Capillary Blood Glucose Levels ( mmol/l)
• HbA1C %
• Hospitalization rate
• Body Mass Index
Key Performance Indicators
SMS Nurse/Doctor Evaluation ReminderThe doctor/health care worker records the frequency at which the patient should come for evaluation. The system then subsequently reminds the patient at appropriate intervals to ensure higher conformance to the visit schedule.
SMS Location and Source of EducationalMaterial : When educational material becomes available, the system informs the patient via SMS about the nearest locations where that material can be accessed.
Collaborate via Social Networking Portal : Patient, Doctors, Nurses, etc collaborate on a social networking portal to share information, concerns etc, and to post queries and answers.
SMS Medication Reminder: The doctor/health care worker records the frequency at which the patient should take medication. The system then subsequently reminds the patient at the requisite periodic intervals to ensure higher conformance to the medicine schedule.
:
Gauteng Department of Health (GDoH) Prototype Overview
:
• Blue IQ (orchestrator)
• GDoH (healthcare
expertise)
• Doctors, pharmacists,
patients (participants)
• Content providers, ISPs,
1 3
2 4
33
Stocks, flows and their causal relationships. Structure as interacting feedback loops
The SD Model as a candidate to help optimize Compressed Morbidity
Adoption from Advertising
Adoption from Word of Mouth
Total Population
Adoption Fraction
ContactRate
Advertising Effectiveness
Potential Adopters
Adopters
Adoption Rate
B
B
R
+
+
+ +
+
+ -
+
Bass Diffusion model in VenSim
Strategic Applications
34Strategic Applications
Applying the basics of SD to the Diabetes Opportunity
Population
Incidence Rate
Un-diagnosed
DiagnosedManaged
DiagnosedUn-managed
DiabetesMortality
Diagnosed & Managed Adoption
Death Rate Managed
Diagnosis Rate
Death Rate Un-managed
Managed Adoption
35
One of the major findings of the prototype was a clearer understanding of what it will cost to drive digital inclusion across Gauteng
System Dynamics Economic Model Overview
1
2
3
4
4
Macro view of model and the four major sections Population & Internet
Adoption
2 Diabetes patient lifecycle
3 Four management activities
4 Six major complications
1
• Inflow of diabetes patients
• Lifecycle from
undiagnosed, through
diagnosis ending in
mortality
• Diet
• Exercise
• Self Management
• Clinic Management
• Visual complications
• Cardiovascular problems
• Amputations
• Neuropathy (Nerve)
• Nephropathy (Kidneys)
NOTE: Variables from Gauteng, South Africa and American diabetic sources combined with prototype-specific findings
Strategic Applications
36
One of the major findings of the prototype was a clearer understanding of what it will cost to drive digital inclusion across Gauteng
Digital Inclusion: Can the right app really bridge the digital divide?
Technology adoption shows
some departure from the
usual curve, with more
people in incent and
mandate category
Costs ~ ZAR1,100 per
person to adopt technology
It should cost ~ ZAR1.1 Bn
to make Gauteng fully
digitally included
Strategic Applications
37
Attempts were made to broaden social circles and consequently make the
participant‟s worlds a bigger place. Over 60% of registered users were
active in social media
Social Inclusion
Source of
Information No. of Respondents
Interactive
media
52 patients mentioned online forum and
blogs as important sources of
information. Therefore, we do see that
people are expressing interest in being
socially connected through ICT
Traditional
Media
201 patients mentioned Traditional mass
media (e.g. newspaper, TV, radio) as
source of info for Diabetes
Doctors &
Nurses
146 patients indicated that they still trust
doctors and nurses more and would go to
them for any information
Friends &
Family
85 patients mentioned that they would
reach their family members and friends
for diabetes related info
Interactive Media
Traditional Media
Doctors & Nurses
Friends & Family
Strategic Applications
38
Patients showed a strong proclivity to adopt the Internet as a means of education and information gathering
Strategic Applications
Technology Adoption Overview
Patients‟ response to simultaneous exposure to three
different technologies – Internet, Mobile, Medical Device
Segmentation
High usage of all three
technologies i.e. internet, mobile,
and Glucometer
Moderate users of all three
technologies
Intervention users who didn‟t use
Glucometer but used internet
and mobile device
Group who used Glucometer
and mobile device
39
Relationship between higher website usage and improved health conditions/awareness
Strategic Applications
Service Health Findings
40
Mobile Devices were pivotal in increasing the hospital appointment compliance
Compliance is critical in managing Diabetes, here the pilot excelled
Snapshot Appointment Compliance Data for July and August
Strategic Applications
41
Complications Population Expense** ZAR
Ketoacidosis 8 16,520
Visual 211,778 599,246,177
Amputations 6,813 42,839,276
Neuropathy 20,501 58,010,938
Cardiovascular 89,599 244,140,486
Nephropathy 6,384 18,065,331
TOTAL 335,083 962,318,728
Complications Population Expense** ZAR
Ketoacidosis 17 34,636
Visual 442,484 1,252,053,618
Amputations 14,202 89,303,542
Neuropathy 43,281 122,467,075
Cardiovascular 101,779 277,326,773
Nephropathy 7,252 20,520,972
TOTAL 609,015 1,761,706,616
Diabetes Complication Incidence with
NO INTERVENTION*
Diabetes Complication Incidence with
BROADBAND ACCESS & SERVICES
INTERVENTION*
~R800 million
Estimated cost
savings of Diabetes
hospitalizations for
Gauteng:
=
R524 Average Cost per
Inpatient Day
~1.5M Hospital Days
Saved
x
INTERVENTION IMPACT
Note: (*)Based on 6 year modeled impact; (**)Expenses were calculated using a unique average
length of stay for each complication
Source: 1Estimating the Cost of District Hospital Services, Joseph Wamukuo & Pamela Ntutela
1
Strategic Applications
Sizing the Opportunity through SD Modeling
42
This prototype illustrates the economic benefit per capita from just one
service. The ~$90/citizen cost of adoption could be spread over multiple
services to maximize the benefit.
Overview of Costs and Potential Benefits
~$90/citizen to deliver
and have services
executed
-
~$230/citizen of
realized benefit
=~$140/citizen for
just one critical
service*
* It is highly likely that one citizen will realize
benefit from multiple services
Source: ICT Enabled Preventive Intervention, Diamond Consultants
Strategic Applications
43Strategic Applications
Two Case Studies
Diabetes Management Healthcare Policy (UK)
44
Model of UK Health and Social Care – NHS, Primary Care
Trusts, Local Government Social Services Directorates
System dynamics model of a typical health community
covering the whole patient pathway from primary care,
through hospitals and onward to post-hospital services
Incentives and penalties in one part of the chain can lead to
„coping‟ policies that can be counter-productive
Based on work carried out by Eric Wolstenholme and others
in UK (1999-2007)
Strategic Applications
Modeling „Coping‟ Policies in UK Healthcare Systems
Source: Coping but not coping in health and social care: masking the reality of running organizations beyond safe design capacity;
Wolstenholme et.al. Vol 3. No. 4, System Dynamics Review, 2007
45Strategic Applications
Patient Flow across Primary Care, Hospitals, and Social Care
Source: Coping but not coping in health and social care: masking the reality of running organizations beyond safe design capacity;
Wolstenholme et.al. Vol 3. No. 4, System Dynamics Review, 2007
46
Delayed hospital
discharges started rising
rapidly
The government felt that
Social Services could do
much better at assessing
and placing older people
in post-hospital services
Fined Social Services for
delayed discharges
Problem started getting
worse – Why?
Strategic Applications
Situation – Delayed Hospital Discharges Rising
Delayed Hospital Discharges
Source: Coping but not coping in health and social care: masking the reality of running organizations beyond safe design capacity;
Wolstenholme et.al. Vol 3. No. 4, System Dynamics Review, 2007
47
When this structure was simulated over 3 years the results showed
significant accumulations in the “medical treatment backlog” and
“waiting discharge to post-hospital services” states, over those observed in practice –
even though they were not allowed in practice.
Flow of medical inpatients and capacity structure of Hospitals and Post-Hospital services
Source: Coping but not coping in health and social care: masking the reality of running organizations beyond safe design capacity;
Wolstenholme et.al. Vol 3. No. 4, System Dynamics Review, 2007
Strategic Applications
48
Formal policies were being „overridden‟ by informal policies that had an adverse impact on the overall flow of patients through the system
Medical Inpatient Model with four „Coping‟ Policies
Source: Coping but not coping in health and social care: masking the reality of running organizations beyond safe design capacity;
Wolstenholme et.al. Vol 3. No. 4, System Dynamics Review, 2007
Strategic Applications
49
Informal Policy: Length of stay in
hospital for normal cases became
a managerial policy variable,
rather than a constant based on
patient need and condition
Positive Impact: Early discharge of
normal patients is an effective
option for hospitals to reduce their
medical treatment backlog
Negative Impact: Reduced length
of stays in hospital create
incomplete episodes of care and
this can result in increases in the
percentage of readmissions.
Strategic Applications
„Coping‟ Policies – Early Hospital Discharge
Source: Coping but not coping in health and social care: masking the reality of running organizations beyond safe design capacity;
Wolstenholme et.al. Vol 3. No. 4, System Dynamics Review, 2007
50
Informal Policy: Transfer of medical
patients to surgical beds whenever
referrals exceeded bed capacity
Positive Impact: Reduces
immediate medical treatment
backlog
Negative Impact: Medical patients
occupying surgical beds result in
cancellation of surgical procedures
and increase in elective surgical
wait times
Conditions of patients waiting will
deteriorate and cause medical
emergencies, and push the
medical treatment backlog
Strategic Applications
„Coping‟ Policies –Overspill of Medical Patients to Surgical Beds
Source: Coping but not coping in health and social care: masking the reality of running organizations beyond safe design capacity;
Wolstenholme et.al. Vol 3. No. 4, System Dynamics Review, 2007
51
Informal Policy: With excessive
waiting for medical admission to
hospital, the referral threshold was
changed to reduce referrals
Positive Impact: Reduced
immediate medical treatment
backlog
Negative Impact: Pushes demand
further back upstream and
ultimately this has to be absorbed
by stocks outside the health and
social care system
Strategic Applications
„Coping‟ Policies – Service Referral Rate
Source: Coping but not coping in health and social care: masking the reality of running organizations beyond safe design capacity;
Wolstenholme et.al. Vol 3. No. 4, System Dynamics Review, 2007
52
Insights: In an attempt to suppress demand and accelerate throughput,
coping mechanisms (fixes) are put into place that may do more harm
than good, by impacting people (inside and outside of the organization‟s
boundaries) in such a way that they do not get the care they need,
although the organizations existing metrics might not tell you that
Solution: Increasing the care package capacity within social services
was not only shown to be a cheaper solution than increasing hospital
capacity, but was demonstrated to be a win–win situation for both health
and social services
Strategic Applications
„Coping‟ Policies – Insights
Source: Coping but not coping in health and social care: masking the reality of running organizations beyond safe design capacity;
Wolstenholme et.al. Vol 3. No. 4, System Dynamics Review, 2007
53
System Dynamics is an effective way of modeling healthcare policies at
the
– Patient level
– HMO, PPO, POS level
– National levels
It can model formal and informal policies and behaviors of all
stakeholders
Effective way of combining statistical data and qualitative information
Simulate behaviors and delayed feedbacks over time
Strategic Applications
Summary
54
Market Context
Our Predictive Modeling Thesis
Strategic Applications: System Dynamics Modeling &
Demonstration
Operational Applications: DRIVE
Behavioral Economics in Healthcare
Questions
Content
Content
55Operational Applications
Strategic Vs. Operational Predictive Modeling Tools
Analytical TechniquesPrediction e.g., Linear & Logistic
Regression
Segmentation e.g., CHAID & Factor
Analysis
Optimization e.g., genetic algorithms
and linear programming
Deterministic
Predict equilibrium point
Linear flows
Point solutions
E.g., claims fraud, segmentation &
targeting, efficacy of disease
management program
Operational Decisions
Simulation TechniquesDiscrete-event Simulation
Agent Based
System Dynamics
Dynamic Systems
Learn drivers of
stock & flow over time
Feedback loops
Systemic understanding
E.g., Disease epidemiology, patient
flow models, impact of public policy
reform
Strategic Decisions
56
Best of breed technology infrastructure
Complements Diamond‟s management consulting practice
Helps clients develop and test predictive modeling prototypes rapidly
Operational Applications
DRIVE Platform : Accelerating Predictive Modeling Solutions
Data Aggregation AnalyticsVisualization &
Reporting
Internal Data
External Data
Syndicated Data
Survey data
Predictive Modeling
Clustering & Segmentation
Optimization
Graphing
Mapping
Lists & Scores
Pivots
Diamond‟s DRIVE Platform
57
Identify patient population & physician group to target prescription drug
compliance and adherence program
Pharmaceutical major trying to move away from a retail detail
model to a more consultative model for marketing to physicians
(also relevant for payors and PBMs)
Example 1
Predict patient population and/or physicians who have patients who are
Operational Applications
58Operational Applications
Marketing Analytics Application Suite - Demonstration: Physicians Map
Source: DRIVE Demonstration; Diamond Analysis
59Operational Applications
Marketing Analytics Application Suite - Demonstration: Output Dashboard
Source: DRIVE Demonstration; Diamond Analysis
60Operational Applications
Marketing Analytics Application Suite - Demonstration: Socio-Demographic Charts
Source: DRIVE Demonstration; Diamond Analysis
61Operational Applications
Marketing Analytics Application Suite - Demonstration: LifeStyle Behavior
Source: DRIVE Demonstration; Diamond Analysis
62Operational Applications
Marketing Analytics Application Suite - Demonstration: Health Risk Factors
Source: DRIVE Demonstration; Diamond Analysis
63Operational Applications
Marketing Analytics Application Suite - Demonstration: Patient Persistence & Compliance
Source: DRIVE Demonstration; Diamond Analysis
64
Predict where to deploy the underwriting resources in the small
business segment of a payer
More RFPs but limited underwriting bandwidth. Need for underwriters to
focus on accounts with maximum likelihood to win and most profitable
Example 2
Identity opportunity and attractiveness of prospective clients and markets
Operational Applications
65Operational Applications
Input
Source: DRIVE Architecture; Diamond Analysis
66Operational Applications
Market Potential Analyzer
Source: DRIVE Architecture; Diamond Analysis
67Operational Applications
Profitability and Ease of capture
Source: DRIVE Architecture; Diamond Analysis
68Operational Applications
Compare Opportunities
Source: DRIVE Architecture; Diamond Analysis
69
Market Context
Our Predictive Modeling Thesis
Strategic Applications: System Dynamics Modeling &
Demonstration
Operational Applications: DRIVE
Behavioral Economics in Healthcare
Questions
Content
Content
70
The current "Information Overload and Accessibility" is resulting in
abdication from decisions to change for both:– Patients
– Providers
Behavioral Economic structured interventions in the information based
decision tree can yield positive results
Many initiatives on individual therapeutic classes have demonstrated
success– Diabetes
– Asthma
– Smoking
At Diamond we are in the process of refining a total HC model where
initial cost reductions from such programs can yield savings in the $1 to
2 billion range nationally
Behavioral Economics in Healthcare
Estimates for Behavioral Economics to reduce costs are varied
71Behavioral Economics in Healthcare
Simple behavioral interventions can influence what people eat and how much they eat
1. Obesity causes at least 300,000 excess
deaths
2. Obesity in adults resulted in health care
costs of $93 billion in 2002
3. Lifetime costs related to diabetes, heart
disease, high cholesterol, hypertension
and stroke among obese are $10,000
more than the non-obese
OBESITY
31%
15%
<20 yrs 20-74 yrs
1. Placing candies three feet away from one‟s
desk reduced volume of chocolate consumption
by 5 to 6 chocolates a day (Self-control)
2. Subjects provided with a bowl of M&Ms in 10
colors ate 77% more than people given a bowl
with only 7 colors (Visceral effects)
3. Food stamp benefits raise food expenditure
more than an equal amount in cash
(Mental Accounting)
4. Pre-ordered healthy-pack options encouraged
healthy eating by Food Stamp Beneficiaries in
Connecticut and North Carolina (Defaults)
5. Having more unhealthy choices reduces the
chances of health options being selected –
Salad, Hamburger, Cake vs Salad and
Hamburger (Choice Relativity)
BE Interventions
Source: Could Behavioral Economics Research help improve Diet Quality for Nutrition Assistance Program participants, USDA,
Economic Service, Diamond Analysis
72
Source: Diamond Retirement Study, 2008
Aspirants
31%
(56yrs/
$50K)Affluent
Sophisticates
24%
(62yrs/
$98K)
Retired Settlers
15%
(66yrs/
$50K)
Survivors
10%
(57yrs/
$24K)
Moderates
20%
(57yrs/
$31K)Percent of
population
Avg. Age/
Avg. Income
High Financial Confidence
Low Health Consciousness
Low Financial Confidence
Low Health Consciousness
Low Financial Confidence
High Health Consciousness
High Financial Confidence
High Health Consciousness
Diamond's market research on baby boomer health and wealth attitudes and behaviors identified five significant clusters of consumers
Diamond has used Agent Oriented Behavioral Modeling on the Baby Boomer Segment
Behavioral Economics in Healthcare
73
The five segments are clearly differentiated in terms of their health consciousness
(e.g., regular exercise, health insurance cover, health risk during retirement)
Agent Oriented Behavioral Modeling
Affluent
SophisticatesAspirants Retired SettlersSurvivors Moderates
Increasing Health Consciousness
15% 27% 29% 30%49%% who exercise at least 3 hours a week
17% 23%50% 60% 84%
% who strongly agree that they have adequate health insurance
26%42% 39%
63% 71%% who ranked physical health as most at risk during retirement
Source: Diamond Retirement Study, 2008
Behavioral Economics in Healthcare
74
Market Context
Our Predictive Modeling Thesis
Strategic Applications: System Dynamics Modeling &
Demonstration
Operational Applications: DRIVE
Behavioral Economics in Healthcare
Questions
Content
Content
75
Changing healthcare landscape
Explosion of Information
Increase in computing power
Emergence of sophisticated tools and techniques
Opportunity to design and model new marketing and behavioral
interventions in healthcare
– DRIVE in Pharmaceuticals and Payer
– System Dynamics in Diabetes Intervention and Policy Formulation
– Behavioral Economics in healthcare
Summary
Summary
76Summary
Diamond Management & Technology ConsultantsPapers and POVs
77
Q&A
Contacts
– Anand Rao ([email protected])
– Richard Findlay ([email protected])
– Amaresh Tripathy ([email protected])
Q&A