Print Paugh Jim

Use of Predictive Modeling
for Claims Management
State of the Industry

Leading, world-class claims organizations implement organizational capabilities that position
them to achieve breakthrough performance.
LEADING EDGE
LEGEND
• Virtual Claim File
Technology
Breakthrough • Predictive Modeling
Process
• Business Rules Mastery
Organization/Human
Capital MIDDLE OF THE PACK • Effective Talent Management
Organizational Capabilities
• Web enabled front-end •• Web

Web enabled
enabled front
front end
end
• Improved Supplier Network •• Improved
Improved Supplier
Supplier Network
Network
• Business process management •• Business
Business process
process management
management
including imaging and rules including
including imaging
imaging and
and rules
rules
• Standardized claims policies & •• Standardized
Standardized claims
claims policies
policies and
and
Table processes, documented & processes,
processes, documented
documented and and
Stakes applied applied
applied
• Established recruiting, training, •• Established
Established recruiting,
recruiting, training,
training, and
and
TRAILING and retention initiatives retention
retention initiatives
initiatives
• Basic systems •• Basic

Basic systems
systems •• Basic
Basic systems
systems
• Few linkages externally •• Few
Few linkages
linkages externally
externally •• Few
Few linkages
linkages externally
externally
• Manual processes; low •• Manual
Manual processes;
processes; low
low degree
degree of
of •• Manual
Manual processes;
processes; low
low degree
degree of
of
degree of automation automation
automation automation
automation
Basic
• Basic human resource •• Basic
Basic human
human resource
resource programs
programs •• Basic
Basic human
human resource
resource programs
programs
programs in place in
in place
place in
in place
place
Low Medium High
Claims Effectiveness
Copyright © 2007 Deloitte Development LLC. All rights reserved. Confidential and Proprietary 1
1
Challenges Facing Claim Organizations
Claim organizations must navigate a difficult course to meet their strategic,
operational and financial objectives.
Prospectively managing your Seeing your trends before poor

book of business losses tell the unfortunate story
Identifying high exposure claims and The right resources on the right claims,
assigning them to the optimal resource taking the right actions
Flawlessly executing the fundamentals, one

Consistency in service delivery
claim at a time
Protecting the balance sheet from

Accurately projecting claim exposure
the ground up
Mitigating inflated loss payouts related to It is estimated that 16 to 30% of insurance

hard and soft fraud payments are related to “soft fraud”
Attracting and retaining capable claims Over the next 10 years, there will be a
handlers despite a tight labor supply projected deficit of 87,000 claim adjusters
Leading Edge Capabilities

Predictive Modeling turns data into information, converts information into insight,
and applies insight to your claim operations to drive claims excellence.
Leading Edge Capability: PREDICTIVE MODELING
• The application of data mining techniques and algorithms to produce a mathematical model that can effectively
predict and segment claims by level of exposure
• The finalized model generally contains 75 or fewer variables from diverse data sources
Description
• At claim intake, model output identifies those claims likely to have above average costs relative to their
expected baseline
• Model scoring can be updated in real time as claim developments or additional variables impact exposure
• The model identifies the segment of claims that disproportionately drives loss costs (20/80 rule)
• Enables smart deployment of resources as the most explosive claims can be assigned to most qualified
Value adjusters
• Provides claim reserving guidance
• Triggers quick business rule application and adjuster intervention
• Integration of model mechanics into existing technology platform

Challenges • Establishing data feeds from external sources
• Effective linkage to business rules and proven claim practices that maximize impact of the model’s output
Claims projected to
have high exposure
Below average claim exposure

2
The Vital Few
Back Injuries
Costs
Cut
Fingers
Type of Injuries
Back Injuries
High
Low High
Low Exposure
3
Pinpointing Costs: Sprain of Back
There is a wide and varying distribution of claim outcomes for
identical type of injury coding.
www.spineuniverse.com
Pinpointing Costs through Segmentation

A relatively small percent of claims represent a disproportionate amount of claim
costs.
4
Dimensions that Impact Claim Outcomes
Claims are inflated beyond appropriate levels based on two dimensions: medical and
behavioral. Identifying claim risk factors provides meaningful insight on how to most
effectively assign and manage claims.
Medical Factors Behavioral Factors
• Co-morbidity • Financial
• Misdiagnosis • Lifestyle
• Improper treatment • Occupational
• Over-treatment • Societal
• Complexity • Medical orientation
High Exposure Claims: Traditional Approach

First Notice of Loss information is evaluated by a Claims Supervisor when assessing claim
exposure and assigning the claim. Below are three soft-tissue back claims. Which claim is
likely to be most costly?
Carol Bill Joe
• Female • Male • Male

• 38 years old • 32 years old • 42 years old
• File Clerk • Welder • Mechanic
• 1 prior claim • 3 prior claims • No prior claims
• Employed 6 years • Employed 2 years • Employed 3 years
• Network doctor • Out of network Dr. • Network doctor
CAROL BILL
JOE
5
High Exposure Claims: PM Approach
Adding non-traditional data elements from multiple sources adds insight and perspective when
evaluating the potential exposure of a claim.
Carol Bill Joe
• Female • Male • Male

• 38 years old • 32 years old • 42 years old
• File Clerk • Welder • Mechanic
• 1 prior claim • 3 prior claims • No prior claims
• Employed 6 years • Employed 2 years • Employed 3 years
• Network doctor • Out of network Dr. • Network doctor
• Lives 42 miles from job • Lives 4 miles from job •Lives 16 miles from job
• Married • Single •Married
• Working spouse • Lives alone •Unemployed spouse
• 3 children • No children •2 children
• (-) Financial stability • (+) Financial stability •Avg. financial stability
• (-) Physician treatment • Avg. Physician •(+) Physician treatment
patterns treatment patterns patterns
Identifying High Exposure Claims

Predictive modeling identifies different combinations of impact conditions that provide
valuable insight on a claim’s exposure level and the reasons behind it.
Claim Leakage Propensity
CAROL
JOE
BILL
LOW EXPOSURE
Copyright © 2007 Deloitte Development LLC. All rights reserved. Confidential and Proprietary
HIGH 11
6
Concept Introduction
Definitions: Data Mining & Predictive Modeling
Data Mining - a process, which utilizes a

number of mathematical techniques, to analyze
large quantities of internal and external data, in
order to unlock previously unknown and
meaningful business relationships. Data mining
lays the foundation for predictive modeling.
Predictive Modeling - the application of data

mining techniques and algorithms to produce a
mathematical model that can effectively predict
and segment future events
Sabermetrics
“Moneyball”
Statistics
–Batting Average
–Home Runs/Singles/DB/TP
–Earned Run Average
–Pitch Count (100), Balls/Strikes
–Curveballs, Fastballs, Sinkers
–On Base Percentage
7
Expected Future Run Study
• Data from 1961 to 1977
• Expected Future runs scored in an inning, as a
function of the number of outs and the currently
occupied bases.
Data is the Key

By combining internal data with public external data, enhanced
segmentation can be achieved.
Claimant Data External Public Databases Line of Business Data

Claimant Specific Information Geographic / Demographic Product, Coverage & Options
Employment & Personnel Info Check Cashing/Sub-Prime Lending Experience Data
Medical History & Treatments Medical / Pharmacy Records Policy Data
Diagnosis Information Employment Records
Consumer / Behavioral / Lifestyle
Financial
Real Estate
Enhanced Census / Behavioral
Claims Data Litigation Weather
Physician Licensing / Disciplinary
Losses Medical Malpractice Heat/Cold extremes
Frequency Industry Claims Precipitation extremes
Timing/Patterns Crime Statistics Wind/Storms
Settlement Data Event Extremes
Loss Control Data
Fraud/Lawsuit
8
Sample Variable Analysis: Age
100%
80% 76%
61%
60%
Bad 40%
24% 25%
15% 17%
20%
0%
0%
-20%
-21%
Good -40% -37%
-60%
-69%
-80%
Below 25 25 to 30 30 to 35 35 to 40 40 to 45 45 to 50 50 to 55 55 to 60 60 to 65 Above 65
Total
Sample Variable Analysis:

Persons Employed in Construction
25%
20%
15%
Bad
10%
5%
0%
-5%
Good -10%
-15%
-20%
-25%
1 2 3 4 5 6 7 8 9 10
Total Cost Relativity
9
Claim Modeling Mechanics
The model produces a score of 1 – 10 (or 1 – 100) that indicates the future
severity relative to an injury type.
~50-75 Variables
Sample Model Equation
Examples:
w
w11(JobClass)
(JobClass) ++ w
w22(Marital
(Marital Status)
Status) +w
+w33(In-Out
(In-Out Network)
Network) ++
• Claimant age
w
w44(Yrs
(Yrs Employed)
Employed) ++ ww55(Wage-Comp
(Wage-Comp Ratio)
Ratio) ++ w
w66(Clmt
(Clmt Age)…
Age)…
• Marital status
• Prescription drug patterns
• Injury date / time 1 - 100
• Change in physician
• Physician type Score
• In network / out of network
• Years of employment
• Salary category Claims projected to
have high exposure
• Ratio of AWW to comp rate
Average claim
outcomes
Below
average claim
exposure
How the Model Works

Predictive Model output is channeled to a business rules engine that
dictates proven, leading practices be performed by the claim handler,
promoting consistent action, every time.
Medical First Notice Claim

Injury
Treatment of Loss Intake
MODEL SCORE Sample Reason Codes

Carol Windsor OUTPUT Financial Instability
16 Circle Ave. Dr. treatment patterns
Burlington, FL 78
92
Attorney Change
Business Leading
TECHNOLOGY
Rules Engine Claim Practices
3 Point Disability,
Assess Assign Set/Adjust Fraud Subro / Claim
Contact / Medical &
Exposure Adjusters Reserves Detection Salvage Resolution
Investigate Lit. Mgmt.
Lit. Mgmt.
10
Standardization through Business Rules
Business rules apply data analytics and logic to develop automated triggers that instantly
prompt an automated task or leading practice to be performed.
Potential Triggers Sample Business Rules
Predictive Model Claim is assigned to a

output indicates a high seasoned claim adjuster with
exposure score with experience with potential soft-
corresponding “reason codes” fraud cases
ACTIONS TAKEN
RED FLAGS
Claimant, who is married with Claim adjuster takes a

children and a working spouse, recorded statement and
sustains a June injury increases the frequency of
claimant contact
Primary doctor has high Claim is assigned to medical

severity or visit frequency per case manager to communicate
injury type (ICD9 code) with providers and monitor
treatment
Focusing on FRAUD
A deliberate attempt either to stage or invent an
accident, injury, theft, arson or other type of loss
that would be covered under an insurance policy.
Hard
Fraud Example:
Staged accidents that include collusion between one or
more of the following: adjusters, agents, physicians,
physical therapists or lawyers
Sometimes called opportunity fraud, it occurs

when a policyholder or claimant exaggerates or
purposely inflates a legitimate claim, resulting in
higher cost outcomes than would be otherwise
warranted.
Soft Examples:
Fraud • Workers’ Compensation claimant is seeking additional
lost time or medical treatment above and beyond what
is appropriate (i.e., locates a new doctor who will deem
the claimant “disabled” when released to return to work
by his treating physician)
• WC claimant seeks attorney when medical evidence
suggests he is capable of returning to a modified duty
job offered by employer
11
Soft Fraud: A Pervasive Problem
Recent surveys show more than 25% of

respondents think it is acceptable to inflate
insurance claims; even more believe it is
reasonable to do so to recover deductibles
Predictive Modeling in Summary

Internal Aggregate
Claims and
Data Cleanse Data
Test Variables MODEL OUTPUT

External to Determine
Data Optimal Set
Internal Create Algorithm

Business that Maximizes
Data Predictive Quality
Reason Codes:
SCORE - 92
Change in Physician
Score Claims Carol Windsor
Synthetic 16 Circle Ave.
to Segregate Burlington, FL 33456
Variables Exposures
12
Right Resource - Right Claim
Special
Investigative
Unit?
Assign to adjusters
with expertise Hard Fraud?
Claim Leakage Propensity

Assign to less Soft Fraud?
Timely to moderately
identification experienced
of claim risk adjuster
characteristics
allows for Claim
Processing
the proper Center
resources to be
assigned at the
outset of
the claim.
Copyright © 2007 Deloitte Development LLC. All rights reserved.

EXPOSURE
Confidential and Proprietary 24
Managing High Exposure Claims

Single Model = Multiple Applications
The core claim model contains several business applications, immediately

enhancing key claim processes and improving resource ROI
Claim Set / Adjust Medical

Fraud Detection
Assignment Reserves Management
Claim Business
Escalation Metrics
RIGHT CLAIM RIGHT RESOURCE
Loss Cost Reduction – LAE Reduction – Improved Customer Service

13
Right Resource – Right Claim
Conventional performance management metrics provide an incomplete picture in measuring
claim adjuster competencies and results.
How do you currently identify your best claim adjusters?

How do you determine their level of skill by claim type?
Traditional Approach Leading Edge Approach
• Supervisor evaluation Traditional metrics plus:

• Claim audit results • Claim outcomes (cost and duration)
segmented by:
• Performance metrics such as:
• claim type (auto, GL, products)
• closing ratio
• predictive model score range
• 24 hour contact
• injury type
• provider network penetration
• attorney representation / litigated
• Client feedback
• subrogation potential
• Attitude • industry/class of business
Driving Results
Successful business implementation drives results that are both
economic and qualitative.
Process Objectives: Financial Metrics:

– Optimally deploy resources – Reduced claim durations
– Apply the established best practice, – Lower loss cost per claim
every time – Reduced LAE
– Standardize triage through automation – Increased adjuster productivity
– Minimize claim duration balanced with – Increased customer satisfaction
quality of care
– Quickly identify and address soft fraud
propensity
– Manage service providers to positively
impact service quality and cost
– Reserving accuracy and timeliness
– Real-time claim escalation Loss Ratio
– System driven quality assurance Improvement
– Efficient use of Loss Control/Prevention = 3 to 6+ %
Resources
14
Impact of Predictive Modeling
Case Studies
EARLY ID
CLAIM ID
HIGHEST
Loss Ratio
WC of BC
WAUSAU TPA RISK ID
Claim
$90pm/mo Saved
Scorecard
$1-$2MM/mo 3 to 6
1.5% savings
points
FRAUD
CANDIDATE UTILIZATION
DETECTION
ID REVIEW
76% review Workers’
Saved Reduced cost
of “bad guys”
$900pp/mo
of bottom
11%-20%/mo Compensation
Med 35% Loss Ratio
15% of cases
Improvement
STAFFING LUMP SUM HIDDEN
19% staff SETTLEMENT NUGGETS
10% turnover Better Reserve $5MM
funneled off 23% Fact Driven Recoup of
of claims Arguments Overpayments
Deloitte Consulting, LLP

185 Asylum St.
Suite 3200
Hartford, CT 06103
Jim Paugh
Tel: 860-725-3175
jpaugh@deloitte.com
Member of
Deloitte Touche Tohmatsu
15
About Deloitte
Deloitte refers to one or more of Deloitte Touche Tohmatsu, a Swiss Verein, its member firms and their respective subsidiaries and affiliates.
Deloitte Touche Tohmatsu is an organization of member firms around the world devoted to excellence in providing professional services and
advice, focused on client service through a global strategy executed locally in nearly 150 countries. With access to the deep intellectual capital
of 120,000 people worldwide, Deloitte delivers services in four professional areas, audit, tax, consulting and financial advisory services, and
serves more than one-half of the world’s largest companies, as well as large national enterprises, public institutions, locally important clients,
and successful, fast-growing global growth companies. Services are not provided by the Deloitte Touche Tohmatsu Verein and, for regulatory
and other reasons, certain member firms do not provide services in all four professional areas.
As a Swiss Verein (association), neither Deloitte Touche Tohmatsu nor any of its member firms has any liability for each other’s acts or
omissions. Each of the member firms is a separate and independent legal entity operating under the names “Deloitte,” “Deloitte & Touche,”
“Deloitte Touche Tohmatsu,” or other related names.
In the US, Deloitte & Touche USA LLP is the US member firm of Deloitte Touche Tohmatsu and services are provided by the subsidiaries of
Deloitte & Touche USA LLP (Deloitte & Touche LLP, Deloitte Consulting LLP, Deloitte Financial Advisory Services LLP, Deloitte Tax LLP and
their subsidiaries), and not by Deloitte & Touche USA LLP. The subsidiaries of the US member firm are among the nation's leading professional
services firms, providing audit, tax, consulting and financial advisory services through nearly 30,000 people in more than 80 cities. Known as
employers of choice for innovative human resources programs, they are dedicated to helping their clients and their people excel. For more
information, please visit the US member firm’s web site at www.deloitte.com/us.
Copyright © 2006 Deloitte Development LLC. All rights reserved.
16

Print Paugh Jim

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Print Paugh Jim

Uploaded by

Copyright:

Available Formats

Use of Predictive Modeling

for Claims Management

State of the Industry

• Web enabled front-end •• Web

• Basic systems •• Basic

Prospectively managing your Seeing your trends before poor

Flawlessly executing the fundamentals, one

Protecting the balance sheet from

Mitigating inflated loss payouts related to It is estimated that 16 to 30% of insurance

Leading Edge Capabilities

• Integration of model mechanics into existing technology platform

Below average claim exposure

Pinpointing Costs through Segmentation

Medical Factors Behavioral Factors

High Exposure Claims: Traditional Approach

Carol Bill Joe

• Female • Male • Male

Carol Bill Joe

• Female • Male • Male

Identifying High Exposure Claims

Data Mining - a process, which utilizes a

Predictive Modeling - the application of data

Data is the Key

Claimant Data External Public Databases Line of Business Data

Good -40% -37%

Sample Variable Analysis:

Total Cost Relativity

How the Model Works

Medical First Notice Claim

MODEL SCORE Sample Reason Codes

Potential Triggers Sample Business Rules

Predictive Model Claim is assigned to a

Claimant, who is married with Claim adjuster takes a

Primary doctor has high Claim is assigned to medical

Sometimes called opportunity fraud, it occurs

Recent surveys show more than 25% of

Predictive Modeling in Summary

Test Variables MODEL OUTPUT

Internal Create Algorithm

Claim Leakage Propensity

Copyright © 2007 Deloitte Development LLC. All rights reserved.

Managing High Exposure Claims

The core claim model contains several business applications, immediately

Claim Set / Adjust Medical

RIGHT CLAIM RIGHT RESOURCE

Loss Cost Reduction – LAE Reduction – Improved Customer Service

How do you currently identify your best claim adjusters?

Traditional Approach Leading Edge Approach

• Supervisor evaluation Traditional metrics plus:

Process Objectives: Financial Metrics:

Deloitte Consulting, LLP

You might also like