PolyAnalyst

1
PolyAnalyst Healthcare Fraud Detection and
Investigation CapabilitiesSergei Ananyan,
Ph.D.www.megaputer.com
12th Annual Medicare/Medicaid Statistics and Data
Analysis
2
Megaputer Intelligence

• Knowledge discovery tools for business users
• Easy-to-understand actionable results

Data Overload
Useful Knowledge
3
Challenge

• Wealth of data are being captured and stored

Licenses
Claims
Patient data
DATA
Professional
Institutional
Pharmacy
DME
4
Challenge

• Overwhelming volumes of data

• 15,000 diagnoses
• 20,000 procedures
• 40,000 providers
• 5,000,000 patients
• 100,000,000 claims per year

• Investigators need automated fraud detection tools

5
Fighting Fraud and Abuse
DATA
Savings
Fraud detection
Investigation
6
Step 1. Fraud Detection
7
Step 2. Investigation
8
Approaches to Fraud Detection

• Whistle blowers
• Rule-based approach
• Data Mining approach
• Unsupervised search for anomalous patterns
• Supervised model building
• Interactive aggregation, visualization and
  reporting techniques

9
Fraud Detection with PolyAnalyst

• Megaputer PolyAnalyst - dedicated Data and Text
  Mining system for Fraud Detection
• Discovers and helps understand situations
  involving previously unknown fraud schemes
• Carries out an objective, data-driven and
  bias-free analysis
• Is a highly scalable server based system
• Facilitates visual design of reusable analytical
  scripts
• Provides interactive data aggregation,
  visualization and reporting capabilities for
  fraud investigation

10
PolyAnalyst capabilities
11
Two types of PolyAnalyst users
Data Analyst
Decision Maker
Visual analytic scenario
Interactive up-to-date reports
12
MediCop
TM

• MediCop detects instances of potential fraud and
  abuse in claims
• Can detect the following types of anomalies
• Inflated prices of individual medical procedures
• Unreasonable substitution of some medical
  procedures with more expensive procedures
• Unnecessary procedures performed
• Unreasonable substitution of some medical
  procedure modifiers with more expensive procedure
  modifiers
• Unnecessary procedure modifiers

13
Fraud Detection MediCop
TM
MediCop
Raw data (CMS-15000, UB-92,etc.)
14
MediCop can discover
Fraud Type Description Recommendation Detection Method
Excessively frequent price Provider charges a higher price for a particular procedure much more frequently than lower prices charged for these or other equivalent procedures by peers Procedure cost, Procedure frequency Procedure Cost Analysis
Overpriced procedure Provider charges irregularly high price for a particular procedure, as compared to peers Procedure cost Procedure Cost Analysis
Excessively frequent procedure The procedure is utilized by this provider for treating a particular diagnosis unusually frequently, compared to less expensive procedures utilized by peers Procedure frequency Procedure Analysis
Extra procedure The procedure was determined to be unnecessary since it is practically never utilized by other providers treating the same diagnosis   Procedure Analysis
Inappropriate procedure A more expensive procedure, which is atypical for the considered diagnosis, was used instead of a cheaper and more typical one Procedure Procedure Analysis
Excessively frequent modifier The provider is using this specific modifier with this particular procedure more often than other providers of the same specialty Modifier frequency Procedure Modifier Analysis
Extra modifier Other providers do not use this particular modifier for this particular procedure, and thus have a lower procedure cost   Procedure Modifier Analysis
Inappropriate modifier The modifier makes a procedure more expensive than other modifiers used by peers in similar cases Procedure modifier Procedure Modifier Analysis
Excessively frequent diagnosis The provider reports higher numbers of patients having this particular diagnosis than its peers, treating patients of the same gender and age group Diagnosis frequency Diagnosis Analysis
Extra diagnosis Other providers never report any patients having this diagnosis for the same gender and age group   Diagnosis Analysis
Inappropriate diagnosis The diagnosis is more expensive and is used by peers less frequently then the recommended one Diagnosis Diagnosis Analysis
15
MediCop Analysis Flowchart
16
MediCop Results
17
MediCop Procedure Cost Analysis
18
MediCop Procedure Analysis
19
MediCop Modifier Analysis
20
Investigation Dimension Matrix
21
OLAP Aggregate Risk Score
22
Highest Losses - Overpriced Procedures
23
Highest Losses - Overpriced Procedures
24
AMRW Compare Billings to Peers
25
AMRW Compare Billings to Peers
26
AMRW Evolution of Billings
27
Pre-payment Analysis - Edits
28
EM and Vaccinations together
29
EM and Vaccinations - overview
30
EM and Vaccinations top providers
31
Trends Graph
32
Analysis of Provider Referrals
33
Text Mining on Medical Records
5,000 of such records per case!!!
34
Searching for Anomalies

• Mismatches between medical conditions and
  selected treatment
• Lack of medical conditions justifying the
  prescribed medications
• Lack of medical conditions for ordered durable
  medical equipment
• Mismatches between prescribed medication dosage
  and recorded objective observations
• Inconsistent patient histories
• Etc.

35
Text Mining on Medical Records
36
Text Mining on Medical Records
37
Semantic Thesauri MeSH, SNOMED, etc.
38
Taxonomy Topics of Interest
39
Taxonomy Drill Down to Side Effects
40
Text Mining Results
41
Benefits
Automated analysis of ALL available data to
quickly focus on the most suspicious cases and
providers
42
Select Customers
Government Insurance Financial High Tech Consumer
Products Manufacturing
43
Questions?
Call [phone] or email info_at_megaputer.com
120 W Seventh Street, Suite 314 Bloomington, IN
47404 USA www.megaputer.com