1. EXECUTIVE SUMMARY
Introduction
In 2005, A Canadian Life Insurance Company provided a database of its dental insurance claims to Claim Analytics for the purpose of applying data mining technology to uncover atypical claims activity. This is a summary of the Claim Analytics research project.
Objective
To determine if the Claim Analytics data mining methodologies could produce benefits in the detection of potential dental claim fraud.
Scope
As this was a research project, we chose to limit our analysis to dental services that were provided:
a) In Ontario
b) In 2004
c) By a general practitioner (which excluded all specialist work).
There were approximately 6,000 dentists that had at least one claim meeting our three criteria and our analysis included all of these claims. Of these, there were 1,644 dentists for whom there was sufficient data for us to classify as typical or atypical.
In total, approximately 200,000 dental claims representing close to $40 million billed were analyzed, where a dental claim is defined as all dental procedures performed on an individual during a single visit to the dentist.
Approach
Traditional Approach
Most of the tools available for analyzing dental insurance claims are rules-based. These tools are very strong at identifying transactions that match known types of potentially fraudulent activity.
However, they are limited to:
a) Identifying the types of abuse that are already known.
b) Analyzing at the level of a single claim, in isolation.
Claim Analytics Approach
For this project, we developed a series of analytical tools to expand on the discoveries that can be made by either rules-based analysis or claim-by-claim analysis.
Our approach used data mining tools to analyze tens of thousands of transactions, thereby uncovering patterns that allowed the measurement of how typical or atypical any given transaction was. This approach enabled the discovery of forms of abuse that are neither predetermined, nor already known.