The property/casualty industry is facing a critical convergence of issues. Diminishing experienced staff, more regulation, increasing governmental scrutiny, low investment returns and an ever-continuing soft market are combining to make one of the most challenging business environments the industry has faced in a long time. Insurers are turning to internal savings as a means of staying in the black, and they have launched multiple initiatives to improve operational performance enterprise-wide.
Since the claims side of the business drives the lion's share of operational costs, carriers are targeting it, in particular, for efficiency improvements. Claims is also a major point of direct interaction with customers and plays a pivotal role in client retention. New tools and techniques in the claims environment are being employed to gain significant improvements in loss adjustment expense (LAE), loss expense (LE) and customer service levels.
Historically, many insurance carriers handled claims in a more or less linear fashion. This approach takes a quick look at the incoming claims and assigns them, in whole or in part, to the best claims handling resource based upon how the claim appears at the time of loss report. That assignment is based at least partly upon workload. As the claim continues down the handling path, the same adjuster stays with it. This general approach does not adequately take into account the tremendous variability in claims complexity and the potential that complexities can develop along the way.
More recently, carriers have invested millions of dollars in establishing "fast-track" units to separate out low-risk claims and handle them using streamlined processing. Applied to the right claims, these fast-track units reveal the benefits of claim segmentation and specialized handling. In principle, the claim would be transferred to more specialized handling if it were to become too complex for the fast-track method, but that requires recognizing in a timely fashion that a claim is no longer appropriate for the process in use. Claims can evolve in unexpected ways. What is required is a dynamic process that continuously monitors and reassigns the claim to the most appropriate resource as conditions change.
Predictive modeling has been used by insurance carriers on the underwriting side for years to determine whom to insure and at what rate. It has allowed carriers to segregate customers very precisely into risk groups and accurately apply a rate appropriate to the risk. There are currently untapped opportunities for applying this approach to the claims side of the business as well, allowing continuous and better segregation of an insurer's book of claims into different risk groups, which promotes more precise handling.
Predictive modeling on the claims side has the potential to allow an insurer to look at its entire claims book from a risk-based point of view. The carrier doesn't really need a model to identify how to handle the most catastrophic claims or the smallest, simplest ones. These are likely to be handled properly because they are self-evident and their "unknown risk potential" is actually pretty well known. Rather, it is best used to identify claims that may appear straightforward yet have the potential to contain complicating factors, such as fraud, subrogation, legal issues and unanticipated medical exposure. Integrated into the claims workflow, predictive modeling can indicate to an adjuster whether a claim has the characteristics of becoming high-risk or is, instead, straightforward and likely to remain relatively low-risk.
To be most effective, predictive modeling needs plenty of high-quality data. The more information and the earlier it can be introduced, the more likely that the optimal decision will be made. In fact, this is what sets this method apart as superior to straight human evaluation. As more data are introduced into the decision-making process, humans quickly become overwhelmed. Predictive modeling can automatically sort through mountains of data and filter out the "nuggets"—the relevant information—and present that to the adjuster for more informed decision making.
Relevant data must be extracted from claims, policy, and other systems and entered into the model in a unified manner. Data contained in legacy systems have often been inconsistent, complex and difficult to access. Though these issues can be overcome, doing so requires considerable amounts of effort and expense. This problem is being mitigated over time. As more companies migrate to modern, unified systems, the data become more accessible, and the expansion of the use of predictive modeling is becoming more feasible. Challenges do, however, remain. For example, much of the most important and most timely information about the circumstances of a claim is contained in the freeform text of the adjusters' notes. To address this problem, more sophisticated text mining has been developed to extract the relevant information from freeform text and present it to the predictive models.
Even with these advances, the use of predictive modeling in claims has, for the most part, been limited to evaluation of the carrier's own claims and policy data. The next step is to take advantage of as much relevant data as possible. This includes data sources external to the insurer's own systems. Public records on individuals, providers and vehicles, medical bill review and estimate data, cross-carrier claims, policy data and accident data are all relevant sources and can flesh out the profile of a claim as it evolves. Introduction of these external data sources further improves the efficiency and precision of predictive modeling techniques.
In the end, action must flow out of the model. The goal is to get the right claim to the right adjuster at the right time. Model results must take full advantage of modern rules-based claims decision support engines and be integrated seamlessly into the claims handling workflow in order to truly optimize both claim handling resources and ultimate claim results.
Medical Injury Claims
Predictive modeling has particular potential in the case of medical injury claims where the true nature of the exposure is not recognized until late in the claim. Claims sometimes initially appear simple in nature, but after further development, clues and data elements can surface to indicate the potential for an escalation in severity. Though these cases are not frequent, they are very important because of the disproportionate cost to the company. Fortunately, such claims are chock-full of consistent, relevant data from medical bills. A predictive modeling system that uses this information and other, external sources can generate automated alerts when the conditions indicate that the trajectory of the claim is not consistent with current reserves.
Once an alert is generated, an experienced claims adjuster should re-examine the file. Alerts can also focus the adjuster's attention on certain facts, such as uncommon medical treatment patterns. The goal is to provide the smallest number of alerts with the highest relevance and to do this as early as possible in the claim's life cycle.
A key factor to the success of an optimized claims workflow is how it is used by the adjusters and their managers. While the claim is progressing, every piece of information should be entered into the model. A risk score for the claim will be generated based on the initial information gathered, but as other events begin to unfold, every piece of data must be captured. The risk score generated in the workflow is dynamic, and the claim should be constantly re-evaluated as new information is entered into the workflow application. By re-scoring a claim continuously, the adjuster is able to better manage the file and to receive alerts if and when the nature of the claim changes.
Insurance carriers can further optimize their workflow application by setting rules within the system that will automate some of their processes. Claims that are more complex will require more action by the adjuster, more steps to follow and specific actions that the adjuster should complete based on the company's best practices. The goal is not to reduce complexity but to identify and manage the complexity so it can be properly addressed and acted upon.
Optimizing Claims Processes Helps Fight Fraud
In addition to gaining efficiencies and lowering operational costs, optimizing claims processes allows insurance carriers to better detect fraud. Predictive modeling can identify fraud based on patterns found in the information provided, such as unusual or suspicious medical treatment and billing. While some instances of increasing claim severity result from unanticipated developments in the medical treatment of the injured claimant, other instances may result from medical fraud. The risk score generated by the claims workflow can alert an adjuster to a claim with suspected fraud. An experienced claims adjuster can examine the claim and ensure that appropriate action is taken, including tighter monitoring of medical treatment by a nurse case manager or referring the claim to the special investigation unit for further action. By conducting a fraud analysis in real time, an adjuster can examine a suspicious case and conduct further investigation prior to paying the claim, thereby avoiding the need to attempt to retrieve funds that have already been paid.
The Bottom Line
Efficient and effective claims processing is a critical part of an insurance carrier's business and an area where most carriers can make improvements. Predictive modeling enables claims organizations to maximize their efforts and resources, including their adjusters, their data and their workflow application. Optimizing the claims organization allows insurance companies to serve their customers better, gain efficiencies and improve their profitability in the short and long terms.
Ernest Feirer is vice president and general manager for LexisNexis Claims Solutions. Dave Torrence is senior vice president and general manager–Auto Casualty Solutions for Mitchell International.