ARTICLE: Predicting Better Claims Management

Cambridge executive Kevin Turner co-authors an article on predicting better claims management for Risk Management magazine.

Risk managers face many complex decisions when selecting third party administrators (TPAs) to manage their claims. All too often, there are implications beyond claims handling that affect cash flow, reinsurance, employee relations and even regulatory compliance. Yet as these issues loom, new technologies are entering the claims handling market to improve handling and decisionmaking abilities .

For nearly a decade, predictive modeling has been used extensively in property/casualty underwriting, most often in personal lines and more recently in the commercial lines market. Through the adoption of these analytic technologies, carriers are recognizing the value of “right pricing” risks based on hundreds of variables as a means of maintaining a competitive edge. The advanced data collection and mining techniques help TPAs and risk managers maximize claims handling and operational efficiency, while controlling costs and resources (e.g., special investigation units, medical specialists and triage nurses) throughout the life of a claim.

Additionally, predictive modeling can help the risk manager oversee selfinsured programs by reducing the need for “special handling instructions,” identifying the injured workers that require the most assistance and helping distinguish between meritorious and potentially fraudulent claims. This technology can also improve pricing and reserving, allowing risk managers to develop precise funding projections and reduce the frequency and impact of unexpected funding requirements as a result of claim escalation.

Right Resources, Right Time
The ability to quickly and effectively direct and resolve a claim is integral to the claims management process. A Menninger Foundation study found that if a claim is not resolved within 60 days of being filed, the chance of an injured employee returning to work is greatly reduced.

By applying predictive modeling at the point of intake, a TPA gains a better understanding of the claim, allowing it to identify and prioritize appropriate action immediately. In the case of workers compensation, predictive modeling can flag a claim that may require special medical intervention much sooner, helping the injured employee receive prompt and quality care. This speeds up the claims process and increases the employee’s chance of making a full medical recovery and returning to work.

Additionally, such technology allows companies to prioritize a claim commensurate with its risk exposure. Usually, claims professionals evaluate a claim based on personal data such as age, employment status and prior claim history—a practice that often leads to misassigned resources, extended claim duration and increased operating costs.

Predictive modeling combines this traditional information with internal data, adding insight to a claim’s exposure level, which can help risk managers better allocate resources. Modeling can analyze hundreds of risk characteristics based on available data and produce a numerical score—as well as the reasons behind it—indicating a claim’s exposure level and complexity. When applied at first notice of loss, a claim that produces a score of 77 may indicate the need for an expert case manager while a claim with a score of 92 may require the attention of the special investigation unit.

Take, for example, a workers compensation claim filed by a 30-year-old clerk who has been steadily employed with the same company for six years, uses a network doctor and has only one prior claim. Based on the aforementioned data, a claims professional could reasonably assume the claim’s exposure level to be low, particularly if the injury is not serious.

Now take into consideration the following additional data: as part of her recovery, she is over-medicating and receiving services from a provider with a history of over-treatment. By inputting this data into a predictive model along with hundreds of other claim characteristics, the algorithm can more precisely assess the exposure level as moderate to high, which may require a more skilled adjuster. Such prompt action has the potential for substantial savings, both in overall cost and manpower.

By using the predictive model to explore and pinpoint exposure, risk managers can optimize resource deployment and minimize claim duration. Additionally, by assigning a quantifiable score to each claim, adjusters can engage expert resources much earlier in the process and provide essential information about the reasons behind the claim’s score. The resulting intervention can mitigate or even deter malingering behavior, leading to reduced loss costs and loss-adjustment expenses.

Enhancing Fraud Detection
Another benefit of analyzing claims based on complexity and exposure levels is better fraud identification. Typically, the industry has relied on the concept of “red flags” and other judgmental measures to alert claims professionals to potentially fraudulent characteristics, which then prompts a referral to specialists for deeper investigation. This method has several limitations that could result in overlooked fraudulent claims or produce “false-positives” that flag legitimate claims for needless investigation because they match a predefined profile.

One widely used red flag, for example, is the Monday morning injury report. The belief is that there is a greater chance of reporting a fraudulent work-related injury just after the worker returns from the weekend. While it can reasonably be assumed that this is the result of a desire for extended time-off for a non-work related weekend injury, further quantitative examination shows otherwise. In fact, a predictive modeling analysis of closed claims actually categorizes a Monday morning injury report as a “green flag,” or a claim with a low level of fraud risk.

Achieving Better Pricing
The use of predictive modeling also allows repetitive tasks to be automated in order to fast-track low exposure claims. This reduces a claim’s cycle time and, thus, the amount of time an adjuster has to spend on it. For example, a claim with a low severity injury and a low score could be marked for autoadjudication, which minimizes adjuster involvement and allows him or her to focus on those cases that need more immediate attention. In the same manner, a claim with a high score can be assigned to an adjuster with the proper expertise. The advantage is that these assignments occur at the first notice of loss as opposed to weeks or months later.

Predictive modeling also allows risk managers to implement timely and effective reserving practices. By examining the circumstances surrounding claims data, TPAs can identify and analyze loss patterns. The risk manager can set initial case reserves more quickly and accurately and make more informed reserve adjustments as needed.

Risk managers can also treat all claims equally at the point of intake and develop pricing structures based on a claim’s overall exposure and complexity. For instance, there no longer needs to be a distinction between “medical-only” and “lost-time” in initial claims handling. All claims become subject to the model, which will determine whether the lag between the injury date and the first day of disability is a strong indicator of increased exposure. Continuing to treat medical-only claims with limited attention until traditional red flags are evident defeats the underlying premise of the technology and often leads to increased loss adjustment expenses and other claim escalation costs. Modeling allows risk managers to develop pricing structures based on the forecasted risk level of a claim while reducing the overall financial impact of claim escalation.

A Shift in the Claims Landscape
The application of predictive modeling technology to the world of claims is still fairly new. Today, the use of analytic models is typically limited to one aspect of claims management, be it loss reserving, fraud detection or medical management, or specific lines of insurance, such as workers compensation or auto. The technology is evolving, however, and the claims management landscape is already moving toward a more comprehensive approach that can be applied across multiple business processes. By using a single model applied to a claim at the first notice of loss, claims providers will improve key processes such as claim assignment, fraud detection, loss-reserve setting and, ultimately, resource return on investment.

This technology will also help risk managers identify better performing divisions, improve forecasting capabilities and aid in the development of targeted assistance programs ranging from safety training to more efficient medical management.



Click here to download the pdf.

Get the latest news from
Cambridge delivered right to your desktop