1

NHCAA CONFERENCE

Installing Predictive Modeling Claims Scoring

TechnologyNovember 18, 2010

2

Welcome to Highmark Primary Offices in

Pittsburgh & Camp Hill along with satellite offices throughout the USA

18,500 Employees 4.6 Million Health Care

Members 28 Million Members

Across all Products 2008 Revenues: $12

Billion

Strategic Initiative toCombat Health Care Fraud Three Phase Approach

– Enhance query capabilities of existing claims history repositories.

– Purchase a provider scoring model which allows for automated detection of aberrant provider behavior.

– Purchase an automated claims scoring model which scores suspicious claims prior to payment. 3

4

Fraud Detection and Investigation Process Flow

RuleBased Detection

Referrals

Recovered orAverted Costs

Evaluation / Prioritization Investigation

DataToolsPeople

DataToolsPeople

DataToolsPeople

DataToolsPeople

DataPeople

DataToolsPeople

PredictiveModeling

FinancialRecovery

DataToolsPeople

PaymentPrevention

DataToolsPeople

Provider ClaimsReview

Phase I

Phase IIProvider Scoring Model

Phase III -Predictive Modeling Claims Scoring Model

Tools

5

Implementing Pre-Payment Claims Scoring TechnologyCritical Success Factors:

– Efficient data transmission of claims (post adjudication but prior to physical payment).

– Building an effective evaluation team: Appropriate skill sets New job rating/job class Job performance measurements

– Establish effective workflow processes between evaluation team and existing departments

Fraud cases Abuse cases Policy changes System changes

– Define and implement an effective value realization/ROI process

PROJECT ORGANIZATION

6

Executive Sponsor

Business Owner

Steering CommitteeBusiness Owner

Program ManagerProject Manager

Claims Processing VPEvaluation Team

Manager

Program Manager

Governance TeamAudit VP

Program ManagerEvaluation Team

Manager

Pre Payment TeamEvaluation Team

Information Services Group Project

Manager

Information Services Group Team

Information Services Group Resources only needed until data extracts are in place

IMPLEMENATION TIMELINE

7

Hourly Extract Implemented 09/09 Go Live for Testing 03/29/10Initial Steering Committee Meeting

6/9/10

Pre-payment process flow

8

Value Realization Term Definition

TERM DEFINITIONIdentified Savings Monies identified from suspicious providers, contracting,

system edits or medical policy issues identified by the FICO pre-payment software. Calculations for ROI are 12 months.

Un-pursued savings Dollars identified by Pre-Pay but cost prohibited to pursue.Deny Button Savings Calculated each time a team member denies claim

payment.Previously Identified Savings

Dollars identified by the software in which an active case is already addressing the issue.

Savings referred to other plans

Dollars identified by Pre-pay but the case has been transferred to another plan (IBC, NEPA, etc.)

Actual Savings Monies received from refunds, restitutions, settlement agreements, claim offsets, and claim denials.

9

Included in calculation of identified savings

Case Summary

Unique Highmark Cases Opened 100Highmark Cases Opened affecting multiple providers

19

Total Highmark cases opened 259Cases related to un-pursued savings 14Events related to previously identified cases 2Cases referred to other plans 5

10

11

Pre Pay Case Trending

April

1 - M

ay 28 Jun

e July

Augu

st

Septe

mber

Octobe

r0

20406080

100120140

14 15 21 21 16 1525

15

42

127

23 18

Unique CasesTotal Cases

12

Pre Pay Claim Trending

April

1 - M

ay 28

(Test

Perio

d) June Jul

yAu

gust

Septe

mber

Octobe

r0

1,000

2,0001,192 1,219 1,133

1,6612,024 1,816

0 99 97 101 62 154

ReviewDenied

13

Net ROI per month:Identified savings divided by monthly

operational costs

April

1 - M

ay 28

(Test

Perio

dJun

e July

Augu

st

Septe

mber

Octobe

r

ROI to da

te048

12

Hit RateClaims reviewed 9,045

Highmark Cases opened 259Case hit rate 2.9%Deny button hit 513

Deny button hit rate 5.7%

14

15

Case hit rate:Claims reviewed divided by cases created

Test Period June July August September October0123456789

16

Deny Button Hit Rate:Claims reviewed divided by claims denied

June July August September October0123456789

17

Productivity Graph

Other P

roject

sJul

yAu

gust

Septe

mber

Octobe

r0

20406080

100120

Days WorkedVacationSystem OutagesOther Projects

Opened case value by monthly operating costs (MOC)

62

3

10

5 3

Case $ value < 50% of MOCCase $ value 50% - 100% of MOCCase $ value 101% - 150% of MOCCase $ value 151% - 200% of MOCCase $ value > 200% of MOC

18CONCLUSION: The identified savings of approximately 22% (18 of 83) of cases opened exceed the operational costs for an entire month.