Forensics Journal - Stevenson University 2014 | Page 69

STEVENSON UNIVERSITY

program participants who receive benefits or other assistance based on
their submission of falsified information. In essence, these recipients
make fraudulent claims to qualify for payments or services for which
they would not qualify otherwise. This type of fraud is harmful not
only to non-profit organizations, but also to legitimate applicants
who may be rejected because fraudulent recipients have usurped the
designated resources. Since this will always be a risk to non-profit
organizations, understanding related predictors becomes a strategic
tool in identifying and denying fraudulent claims. Predicting external
fraud can also be particularly difficult when any number of variables,
or combination thereof, may be considered key indicators or “red
flags” that a claim may be fraudulent and requires additional inquiry
or analysis to determine its validity. The central purpose for these
predictors is for non-profit organizations to increase internal confidence of identifying fraudulent claims before assistance or benefits are
expended in error. This can be accomplished through the application
of fraud detection models utilizing data mining, where this analytical
approach to fraud detection uses historical data to “identify possible
predictors of fraud associated with known fraudsters and their actions
in the past” (Nisbet et al. 350). Generally speaking, this methodology
utilizes pattern analysis in the creation and application of rules, general principles, red flags, and alerts to generate a decision that recommends acceptance, rejection, or review; or a score or profile indicating
the likelihood of possible fraud. Predictive analytics and link analysis
may be more common end-user terms for some of these fraud detection models. Fraud detection systems are also available commercially
as fraud detection software. The most common application for these
fraud detection models are credit card fraud, check fraud, application
fraud, claim fraud, and healthcare fraud – all forms of external fraud.

Again, the most prominent example of external fraud against a nonprofit organization is healthcare fraud. The Federal Bureau of Investigation estimates $80 billion in annual healthcare fraud costs, and
close to $700 million annually when fraud and improper payments
are combined (Worth). The Centers for Medicare and Medicaid
Services (CMS) is a prime target for fraudulent claims. Processing
more than 4.8 million claims per day, CMS is “the most well-known
healthcare entity using analytics” (Worth). The 2010 Small Business
Jobs Act authorizes the use of “predictive modeling and other analytics technologies to identify improper claims for reimbursement and
to prevent the payment of such claims under the Medicare fee-forservice program” (Roehrenbeck). This predictive modeling program
is intended to capture and provide a comprehensive view of provider
and beneficiary activities across all regions with the goal of detecting
patterns and networks that represent a high risk of fraud (Roehrenbeck). This modeling program can also be used to identify other
improper payments not related to fraud.
The analytics are divided into four categories: normal rules of healthcare and how these rules may be violated; anomalies; predictive
modeling using patterns based on cases of fraud; and social networking that analyzes the financial ties of a fraudster-provider (Worth).
This fourth category is a form of link analysis, which is “a data analysis
technique that examines the relationships among claims, people,
and transactions” (Schreiber). With the natural evolution of fraud
schemes and interrelated factors, the inherent “intelligence in the predictive analytic system ‘learns’ from the new rule patterns and builds
increasingly more sophisticated models,” adapting for new types of
fraud as new rules are developed (Schreiber). Predictive analytics
and link analysis relate a greater number of highly arbitrary, and even
unexpected, variables to detect and prevent fraud. This is significant
as external fraud can be more elusive and less discernible when compared to the typical predictors of internal fraud.

Due to the nature of healthcare needs and associated costs, the rate of
external fraud against non-profit healthcare organizations is probably
one of the highest. A 2004 article notes that the National Health
Care Anti-Fraud Association “estimates 3% to 10% of every dollar
spent on healthcare in the U.S. is lost to fraud, totaling $39 billion
to $150 billion a year” (Mantone). With increasing healthcare costs,
one would only expect these fraud losses to increase. To mitigate this
risk of external fraud, non-profit healthcare organizations currently
use fraud detection software and predictive modeling in a continuous effort to detect fraudulent claims before making payments. The
Government Employees Hospital Association uses a Fair Isaac Corporation (FICO) software product that assigns a fraud-risk score for
each billing transaction. According to the Association’s data analysis
manager, the software also increases productivity as employees can
focus more on the fewer number of flagged claims instead of attempting to review all claims as possible fraudulent cases (Mantone).
Overall, the Association is better equipped to reduce fraud and related
costs, as well as improve employee productivity and positively impact
customer service.

While the use of data mining, predictive analytics, and link analysis
may be useful for external fraud prevention efforts of larger nonprofit organizations, prospective detection may still prove difficult for
smaller non-profit organizations. These entities can purchase more
modest versions of fraud detection software or build their own internal databases and create rules to flag suspicious applications or claims
for benefits or services. Key f