Transcript MCO Meeting 10-25

September 6, 2012
NCDHHS FAMS Overview for Behavioral Health
Managed Care Organizations
©©
2010
IBM
Corporation
2011
IBM
Corporation
What is FAMS?
IBM Fraud and Abuse Management System (FAMS)
•Developed 16 years ago to allow users platform to use statistical scoring to evaluate peer groups of healthcare
providers, and identify behaviors of fraud, waste, and abuse within the population using behavioral analysis.
•FAMS fits into the investigative process in the identification and Research steps
•Implemented in North Carolina in 2010 in many service areas
Identify
Source: If applicable, describe source origin
2
2
Research
Prioritize
Investigate
Resolve
Background
• Why we are here today:
– http://www.wral.com/news/local/wral_investigat
es/video/11126237/#/vid11126237
– http://www.wral.com/news/local/wral_investigat
es/video/11130897/#/vid11130897
– http://www.wral.com/news/local/wral_investigat
es/video/11356150/#/vid11356150
3
Investigators are challenged to find suspicious behaviors that are buried
within the massive volume of healthcare claims
 Payers are under great pressure to
pay claims quickly
 Fraudsters hide “bad” behaviors
amongst the hundreds of millions of
claims submitted annually
 Investigators are overburdened with
case loads and lack the resources and
technology to find fraud fast
 Payers adopted a ‘pay-and-chase’
strategy, pursuing cases based on tips
received through fraud hotlines
4
4
FAMS integrates technology, people and experience
A library of over 8,500
algorithms that are the
basis for specialtyspecific models and
successful
implementations at
over 40 clients
worldwide
Seventeen years of
experience in helping
public and private
payers detect healthcare
fraud, waste, and abuse
Software Assets
& Tools
A powerful analytics engine to quickly
sort through large quantities of data
using efficient algorithms and
specialty-specific models
5
5
Using behavior modeling can
help find suspicious behaviors
faster
Outlier
Detection
Data mining
and
segmentation
Predictive
Modeling
6
6
 Which providers are behaving differently than
others (in a suspicious way)?
 How “good” or “bad” is a provider behaving,
relative to other providers?
 What it is “normal” behavior?
 What are patterns of non-compliant (and
criminal) behavior that I don’t know about?
 If I catch a “bad” provider, how can I find out
who else is behaving like that?
 Are there groups of providers who are behaving
in the same way?
 Which providers are likely to behave “badly” in
the future?
 What are the indicators that a provider’s
behavior is getting “better” over time? “Worse”
over time?
Behavior modeling uses analytical methods to select outliers that they must be
investigated
Behavior
What behavior is being identified?
What data can be used to discover
“behaviors”?
7
7
Population
What are the characteristics and
relationships of those behaving in this way?
What data can be used to identify “who”?
FAMS helps your investigators to pinpoint suspicious claims by using
advanced analytics to identify “bad” behaviors
IBM Fraud and Abuse Management System (FAMS)
Analyzes healthcare claims for mathematical anomalies
Used after claims are paid to enable investigators to focus on the right providers
Can be deployed at a complaint intake level to validate incoming information
How FAMS if differentiates of a
traditional approach?
• It detects multiple behaviors and
schemes simultaneously
• moves analysis from claim level
to provider level
Focus areas for
consideration are
discussed
FAMS analysis techniques are used
to determine who is behaving
differently and how
• Shortens the time to investigate
and recover funds
• Measures fraud scientifically
8
8
Investigators
conduct further
investigation
Reports are reviewed
and actions planned
Providers are ranked
and scored and
categorized
Key Terms
• Feature
– A feature is a measured attribute of a provider, a
feature is a query run against claims data for a
provider, or a feature is a simple calculation of
claims information. A feature is a numeric or
categorical attribute of an entity used in entity
profiling. In a profile, each numeric feature's value
is translated to a score using the scoring
associated with that feature.
• Model
9
– A Model is comprised of groups of features that
Modeling and Profiling
•
•
Models
– Lists of questions to ask of the data for a peer group
– Consists of 50 to 150 questions related to the peer group, that are organized into subgroups by type of
service or hypothesis
Peer Group Analysis Profile
– Answers to the list of questions scored in relationship to the peer group from 1 to 1000
– Where analysis occurs
– Created by running claims for a provider peer group, for a certain timeframe, against a model
FAMS Model
x
Peer Group Claims Data
=
Provider
Recipient
Date
Service
Paid
12345
1001
1/7/2012
90801
$80.00
54856
2376
1/8/2012
H2022
$225.00
97256
1400
1/10/2012
90806
$75.00
10
Analysis Profiles
FAMS Demo Slides
• FAMS features an easy to use, graphic user interface
– Analytics capabilities include:
• Peer group profile visualization and reporting analysis
–
–
–
–
Composite level scoring and ranking
Feature and feature group scoring and ranking
Tracking change over time
Visualization analysis tools
• Claims reporting functions
– Basic claims reporting functionalities
–
-Claim detail extraction
–
-Reporting on combinations of providers, diagnosis, procedures,
recipients, etc…
– Recipient drift reporting
11
Visualization Analysis
•
•
•
•
•
12
Geospatial Mapping
Peer Group Segmentation
Graphing
Charts
Parallel Coordinates