Moving Beyond ‘Pay and Chase’ to Fight Healthcare FraudHealthcare providers that embrace predictive analytics can prevent the payment of false claims before they undermine business performance.

Executive SummaryFinancial losses in the U.S. due to healthcare fraud are estimated to range from $75 billion to a staggering $250 billion a year, accounting for as much as 10% of U.S. health expenditures.1 Experi-ence shows that preventing fraudulent payments is far more cost-effective than post-payment recovery efforts.

This white paper explains why it is imperative to move beyond the historic “pay and chase” model, to a data-driven methodology that utilizes analytics to perform targeted, clinically-relevant fraud identification and prevention. It then describes how such data analysis can be used to detect and help prevent common types of fraud.

Fraud Flourishes Amid Volume, Complexity The sheer volume and complexity of the $2.6 trillion U.S. healthcare system makes combating fraud a daunting challenge. Every year, hundreds of thousands of service and product providers file billions of claims. Medicare alone pays 4.4 million claims a day to 1.5 million providers across the country. Beyond the sheer number of claims to monitor, healthcare insurance payers are outgunned by fraudsters, many of whom have the specialized knowledge on which to base false claims, such as best clinical practices, medical ter-minology and specialized codes.

The lack of a single claims repository also makes it more difficult to detect multiple claims for the same product or service. Fraud detection is yet more complex because most providers bill multiple private and public payers, with each payer having visibility only into the claims it receives and adjudicates.

Examples of provider-perpetuated waste, fraud and/or abuse include:

• Billing for unnecessary medical services, equipment or prescription drugs.

• Billing for services that were not actually provided.

• Up-coding, or inappropriate use of current procedural terminology (CPT) and diagnosis-related groups (DRG) codes to gain higher reimbursement.

• Submitting claims for non-covered persons.

• Scheduling more frequent visits than required to increase reimbursements.

• Billing for services for which the provider is not qualified.

• Duplicate billing for services rendered.

• Payments stemming from kickbacks.

• Keeping overpayments made in error by payers.

• Cognizant 20-20 Insights

cognizant 20-20 insights | january 2014

2

Today’s retrospective approaches do not identify fraud or efficiently prioritize investigative efforts on a timely basis. A shortage of investigative tech-nology, money and staff leads to reactive and frag-

mented fraud management processes. Recovering fraud-ulent payments after the fact is not only more difficult than prevention, but it also requires large amounts of money and staff time.

Meanwhile, member-perpet- uated fraud, waste and abuse are also proliferating. Examples include:

• Falsifying or modifying prescriptions to obtain alternative substances or medications, or more medications than prescribed by the physician.

• Members sharing plan membership ID cards with non-members.

• Members failing to disclose supplementary health insurance coverage.

• Members obtaining unnecessary equipment and supplies.

The Analytics AnswerMany observers, including the American Medical Association, have long suggested that the most efficient way to combat healthcare fraud is through targeted, streamlined analytics. The reactive “pay and chase” model should be replaced by a multi-pronged approach that uses analytics to enable targeted, clinically-informed fraud identification and prevention.

Such a system would streamline and integrate commercial, federal and state-level anti-fraud programs and audits to improve coordination and detection rates. It would build multidimensional predictive models to detect fraud at a very early stage, and develop innovative solutions to reduce costs by minimizing the administrative burden and more efficiently targeting enforcement.

A successful fraud detection system would also stay up to date with continuously evolving fraud and abuse schemes, detect suspicious claims with a high degree of accuracy before they are paid and offer a cost-effective approach to triaging large volumes of claims.

By proactively utilizing analytics, such a system could identify aberrant claims in real time and cross-reference them with claims from other data

sets to flag potentially fraudulent activity. Doing so requires the use of dynamic, configurable pre-dictive modeling tools that support the following:

• Close to real-time pre-payment detection.

• Recognition of non-linear patterns, such as clustering, discriminant analysis and decision tree.

• Linear pattern identification, such as trending and heuristics to “learn” more about which claims patterns are the most suspicious.

Using analytics also requires healthcare fraud models that are designed and tested to identify fraud, waste and abuse indicators, such as:

• Unusual geographic concentration/dispersion of participants.

• High-frequency utilization designs.

• Uncharacteristic practice patterns.

Such advanced predictive analytics can recognize potential fraud patterns that are undetectable using conventional methods, delivering action-able results through automated detection (see Figure 1, next page). These approaches include:

• Stratifying numbers to identify unusual (i.e., excessively high or low) entries.

• Statistical parameters, such as averages and standard deviations to identify outliers that could indicate fraudulent activity.

• Trend charts to identify patterns of potential fraud among data elements.

• Digital analysis to identify unexpected occur-rences of digits in naturally occurring data sets.

• Bridging diverse sources in the healthcare ecosystem to identify the occurrence of identical values (such as names, addresses and account numbers) that should not exist.

• Identical value identification to flag duplicate transactions, such as payments, claims or items in expense reports.

• Link analysis, which establishes relationships among claims, people and transactions by dis-covering related claims in seemingly unrelated instances.

• Comparing summed and totaled charges using control charts to identify outliers and ambiguous claims.

• Validating entry dates to identify suspicious or inappropriate times for postings or data entry.

cognizant 20-20 insights

Recovering fraudulent payments after

the fact is not only more difficult than

prevention, but it also requires large amounts of money

and staff time.

3cognizant 20-20 insights

Key Requirements

Analytics systems that detect and prevent health-care fraud should offer the following features and capabilities:

• Dynamic rules engines as the system foundation, testing each transaction against a predefined set of business rules to detect known types of fraud or abuse and flag suspicious claims.

• Anomaly detection algorithms, based on KPIs for specific tasks or events, baselined with thresholds that alert for potential fraud. Outliers or anomalies should also be identified to indicate new or previously unknown patterns of fraud.

• Predictive modeling that utilizes data mining tools to build models that produce fraud propensity scores. Such modeling uses statis-tical analysis to discover previously unknown fraud schemes based on previously unidenti-fied metrics.

• Social network analysis and multi-enti-ty fraud detection, using social tools and techniques to identify organized fraud activities by modeling relationships between entities in claims. These entities may include locations, service providers, members, addresses or

telephone numbers. These tools can be tuned to display link frequencies that exceed a programmed threshold.

Case Examples

Typical areas where data analysis applications can detect and prevent healthcare include:

• Highlighting billing for medically unnecessary tests.

• Identifying false/invalid/duplicate entries.

• Highlighting excessive use of DRGs that have historically posed a high risk for fraud.

• Identifying excessive billing by a single physician.

• Identifying employee overtime abuses.

• Reporting claims against authorization records for new or terminated employees.

• Identifying multiple payroll deposits to the same account.

• Matching the OIG (Office of Insurance General) list of excluded providers with employee master files.

• Finding kickbacks paid in exchange for referring business by detecting either analytical or control symptoms or anomalies.

• Identifying charges posted outside of permis-sible periods.

Predicting Fraud Before it StrikesAdvanced predictive analytics recognize patterns that would be undetectable using conventional methods, delivering actionable results via automated detection.

Medical Equipment

Members

Facilities

Laboratory

Pharmacy

Providers

Dynamic Profiles

X-rays to exams

Dollars billed vs. peers

Procedure mix vs. peers

Max single-day activity

Variable N

Fraud Propensity

Score Predictive Models

Claim Number Suspicion Score

1442618 9941383514 9601433949 9211945594 8871525206 8431524787 7981466937 7541576751 7101529271 6821549491 6261394739 6201493195 5871526027 545

High

Fraud Potential

LowPredictive modeling compares claims with baselines or thresholds to create fraud propensity scores.

Derive powerful variables

Figure 1

4cognizant 20-20 insights

• Identifying “up-coding” by flagging statistical outliers.

• Analyzing large invoices that lack purchase orders by amount, provider, vendor, etc.

• Validating lists of enrolled employees by comparing them with lists of people receiving health benefits from insurers.

With new fraudulent billing and claims schemes continuously emerging, fraud prevention must consist of a “virtuous cycle,” in which the results of the latest data analysis help to fine-tune the algorithms, rules and models used to detect potential fraud. Such a system relies on ongoing validation and measurement of the latest fraud detection models, drawing on ongoing experience to increase their effectiveness (see Figure 2 and sidebar).

Looking AheadHealthcare fraud and abuse puts an enormous financial burden on a system that is already facing unprecedented pressure from consumers and legislators to contain costs. The magnitude of potential fraud and abuse savings is so great that as more payers recognize and adopt state-of-the-art detection capabilities, those with antiquated fraud management practices will be at a signifi-cant competitive disadvantage.

To meet these challenges, it’s time for health insurers to devote additional resources to predic-tive modeling technology and real-time analytics. By applying these techniques to fraud prevention, they can avoid the excessive costs and uncer-

tainty of recovering fraudulent payments after they have been paid. Combining multiple tech-niques offers the best chance for detecting both opportunistic and professional/organized fraud. A hybrid approach, which integrates knowledge of existing fraud schemes with powerful predic-tive analysis techniques, can help contain system-wide healthcare costs, while giving providers and payers a competitive edge.

Fraud Prevention’s Virtuous CycleAn ongoing analytics-driven fraud detection program can detect and stop more fraud more efficiently over time.

Figure 2

Quick Take

• Challenge: Identify and create new business rules to prevent claims fraud and abuse.

• Analytics applied: Rules and anomaly detection.

• Services provided:

> We established a claims analytics team of medical coders, fraud analytics investiga-tors, data miners and SAS programmers to identify and create new rules to flag fraudu-lent claims.

> We also created a prospective fraud inves-tigation team made up of healthcare pro-

fessionals and medical coders to analyze and investigate possibly fraudulent claims triggered by the rules, as well as the result-ing anomaly detection to block fraudulent claims.

• Results:

> Saved $8 million by identifying fraud before it happened.

> Created 24 new business rules, with annual potential savings of more than $18 million.

Analytics in Action: A Leading Healthcare Services Company

Big Data

Statistics

ANALYTICS

•Result processing •Measurement •Validation•Feedback

• Massive storage• Parallel processing• Cloud access

• Descriptive statistics• Reports

• Dashboards• Orchestrated graphical views• Intuitive representation

• Predictive methods • Pattern analysis • Link analysis • Classification • Clustering

Interpretation& Authentication

Modeling & Discovery

Data Visualization

About CognizantCognizant (NASDAQ: CTSH) is a leading provider of information technology, consulting, and business process out-sourcing services, dedicated to helping the world’s leading companies build stronger businesses. Headquartered in Teaneck, New Jersey (U.S.), Cognizant combines a passion for client satisfaction, technology innovation, deep industry and business process expertise, and a global, collaborative workforce that embodies the future of work. With over 50 delivery centers worldwide and approximately 166,400 employees as of September 30, 2013, Cognizant is a member of the NASDAQ-100, the S&P 500, the Forbes Global 2000, and the Fortune 500 and is ranked among the top performing and fastest growing companies in the world. Visit us online at www.cognizant.com or follow us on Twitter: Cognizant.

World Headquarters500 Frank W. Burr Blvd.Teaneck, NJ 07666 USAPhone: +1 201 801 0233Fax: +1 201 801 0243Toll Free: +1 888 937 3277Email: inquiry@cognizant.com

European Headquarters1 Kingdom StreetPaddington CentralLondon W2 6BDPhone: +44 (0) 20 7297 7600Fax: +44 (0) 20 7121 0102Email: infouk@cognizant.com

India Operations Headquarters#5/535, Old Mahabalipuram RoadOkkiyam Pettai, ThoraipakkamChennai, 600 096 IndiaPhone: +91 (0) 44 4209 6000Fax: +91 (0) 44 4209 6060Email: inquiryindia@cognizant.com

© Copyright 2014, Cognizant. All rights reserved. No part of this document may be reproduced, stored in a retrieval system, transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the express written permission from Cognizant. The information contained herein is subject to change without notice. All other trademarks mentioned herein are the property of their respective owners.

About the AuthorNitin Gangwani is a Senior Associate with Cognizant Analytics. With wide experience applying business consulting and analytics in the life sciences, managed markets and healthcare industries, Nitin has delivered strategic consulting and analytics services in the areas of business analysis, requirements gathering, data warehousing, patient claim analytics, campaign management and solution design. He holds an M.B.A. from the Management Development Institute and a B.E. degree from Netaji Subhas Institute of Technology. Nitin can be reached at Nitin.Gangwani@cognizant.com.

References

• “Statement of the American Medical Assocation: Examining Options to Combat Health Care Waste, Fraud and Abuse,” American Medical Association, November 28, 2012, http://democrats.energycommerce.house.gov/sites/default/files/documents/ Testimony-American-Medical-Association-Health-Healthcare-Waste-2012-11-27.pdf.

• “Fraud Detection Using Data Analytics in the Healthcare Industry,” ACL, 2010, http://www.acl.com/pdfs/DP_Fraud_detection_HEALTHCARE.pdf.

• “Health Care Fraud, Waste and Abuse,” LexisNexis Web site, http://www.lexisnexis.com/risk/health-care/fraud-waste-abuse.aspx.

• Robert Kelley, “Where Can $700 Billion in Waste Be Cut Annually from the U.S.. Healthcare System,” Thomson Reuters, October 2009, http://www.healthleadersmedia.com/content/241965.pdf.

• Ted Doyle, “The Use of Advanced Analytics to Prevent Healthcare Fraud,” UnitedHealth Group/OptumInsight, October 2012, http://www2.mz.gov.pl/wwwfiles/ma_struktura/docs/11_kdsowoz_07012013.pdf.

• “A Review of Healthcare Fraud and Abuse in America,” AAOMS Committee on Healthcare & Advocacy, March 1, 2011, http://www.aaoms.org/docs/practice_mgmt/fraud_and_abuse.pdf.

Footnote1 “Statement of Louis Saccoccio, Executive Director, National Health Care Anti-Fraud Association, on

Improving Efforts to Combat Health Care Fraud,” National Health Care Anti-Fraud Association, March 2, 2011, http://waysandmeans.house.gov/uploadedfiles/socc.pdf.