7 Ways to Fight Back Against Fraud

Insurance companies use a variety of tactics to detect and deflect fraudulent claims

By David West

Although it’s impossible to calculate the exact cost of insurance fraud (estimates range from US$20 billion to upward of US$60 billion in North America alone), there is no doubt that fraud is a huge problem and an ongoing challenge. But it is a crime that technology can help curtail.

If criminal behavior remained static, identifying patterns and flushing out fraudulent claims would be a simple process. Individuals exploiting the system, however, are typically savvy enough to make identification of nonlegitimate claims quite difficult. As a result, a variety of techniques have been developed to help reduce the number of false claims that insurance providers cover.

Method No. 1: Voice stress analysis
Fraud begins with a lie. Criminals tend to be proficient liars, so a “story” may sound like a fair and reasonable claim. Though words can be controlled and may ring true, certain vocal patterns indicate when a claimant is being dishonest. Voice stress analysis can measure these pattern variations.

Voice stress analysis involves using electronic recording devices to evaluate levels of stress during verbal interviews. As a claimant explains a situation, a recorder compares vocal patterns to a predefined stress threshold.

An advantage of this technology is that stress levels are neutral in terms of gender and ethnicity. The technology also factors out unfamiliar accents, which may complicate an individual evaluation. The test simply measures stress levels, thus reducing human bias.

On the other hand, the technology requires training to use and cannot prove that the claimant is lying. Voice stress analysis is a relatively new process, and detailed results are not publicly available. Use of voice stress analysis may also have a negative impact on customer retention. Announcing that a company uses it may reduce claim padding, but it may offend honest claimants at the same time.

Method No. 2: Red-flagging
At the opposite end of the technology spectrum is the more common use of “red flags.” In this system, adjusters look for specific patterns of activity and flag claims that look suspicious. More experienced adjusters then investigate these red-flagged claims in depth.

The advantage of the red-flag approach is its simplicity. Practitioners and regulators can easily match a list of behaviors to accounts, and minimal investment or training is needed to implement a basic rule-based red-flag program.

Unfortunately, a nonautomated red-flag system has many disadvantages. For one, it puts the burden of detection on overworked adjusters. For another, diligent adjusters often waste time investigating the high number of red-flag claims that turn out to be false positives. Furthermore, flagging rules are based on past fraud experiences, which limits the ability of this approach to detect new fraud techniques.

Method No. 3: Predictive modeling
In recent years, many insurers have turned to predictive modeling processes to help them detect fraudulent claims. This has reduced the need for tedious hands-on account management. Relying on statistics, sage analysts use data mining tools and build programs that produce fraud propensity scores. Adjusters simply input data, and claims are automatically scored and available for review.

Predictive models tend to be more accurate than other methods. Information can be collected and cross-referenced from a number of sources. This method provides a better balance of data than the more labor-intensive red-flag system.

But the initial setup with predictive models is critical. Models are only as reliable as the data on which they are constructed. If fraud classifications are not accurate, then results are of little value. Even with good information, accounts may show false positive results for a variety of reasons. 

Additionally, model performance deteriorates with age. As criminals adopt new approaches, analysts must update programs to reflect new patterns. Yet in spite of some flaws, predictive modeling shows great promise for fraud detection.

Method No. 4: Database searching
Database searching, another approach, would be better classified as an investigative tool. Here, companies subscribe to services offered by various vendors. Subscribers submit skeletal data of adjudicated claims and then have access to data submitted by other members of the service.

The huge bank of collective data powered by such search interfaces enables adjusters to view massive amounts of information from numerous sources. Finding claims associated with search criteria, however, does not necessarily result in fraud detection. Adjusters must be skilled at reviewing and interpreting data to effectively use these services.

Unlike the first four methods, the remaining three fraud detection techniques involve a retrospective analysis of adjudicated claims. The first four, if employed correctly, can prevent payments. The last three techniques are useful for identifying the activities of fraud rings, internal fraud and leakage.

Method No. 5: Exception reporting 
Exception reporting is the most basic approach and the easiest to implement. Key performance indicators associated with tasks are baselined and thresholds set. When an event exceeds the threshold for a particular measure, the system reports the event.

Once in place, the system functions automatically. It monitors adjuster activities and helps identify and correct problems. This is a useful tool for evaluating individual performance and identifying employee training opportunities.

Determining what to measure, what time period to use and what threshold to set can be difficult. Still, exception reporting is an effective tool for internal management.

Method No. 6: OLAP reporting
Another antifraud tool is the combination of ad hoc query and online analytical processing (OLAP) – facilitated by the creation of databases summarized across many different dimensions. OLAP reporting enables analysts to search through huge volumes of adjudicated claims, make comparisons, identify exceptions and find unusual situations in a dynamic environment.

OLAP capabilities are available from many different vendors. An experienced analyst can use the data to quickly generate reports that identify potential problems and direct future investigations more effectively.

Because OLAP is interactive, it requires intervention from an analyst who must have a strong understanding of the data. The creation of OLAP databases is not a trivial task. The database architect must understand the claims process and must identify the required dimensions.

Method No. 7: Link analysis
The final fraud detection tool is link analysis, which is effective in identifying organized fraud activities. Link analysis is the process of modeling the relationships between entities in claims data to uncover abnormal patterns. Entities may be defined as locations, service providers, telephone numbers and Vehicle Identification Numbers – to name just a few.

Tools can be tuned to display link frequencies that exceed a programmed threshold. Large volumes of seemingly unrelated claims can be checked, and then patterns and problems may be identified. Link analysis is relatively automated, but a skilled analyst is required to put all the pieces of the puzzle together.

Insurers have been able to identify the presence of fraud rings using link analysis. Because the focus is on retrospection, this process does not necessarily prevent fraud. It does, however, allow the industry an opportunity to detect organized criminal activities and to take appropriate action.

Companies can use these seven methods individually or in combination to help them detect and prevent criminal claim activities. Employing even some of these tactics may deter would-be fraudsters from making false or padded claims.

Fraud cuts profits, and lax fraud identification practices put a company at a competitive disadvantage. Although it does take commitment and an investment of resources and time to address fraud issues, the stakes are too high to ignore the problem.

*This article originally appeared in National Underwriter Property & Casualty magazine.

Bio: David West is the Insurance Practice Research Director for TowerGroup.