BI’s Article search uses Boolean search capabilities. If you are not familiar with these principles, here are some quick tips.
To search specifically for more than one word, put the search term in quotation marks. For example, “workers compensation”. This will limit your search to that combination of words.
To search for a combination of terms, use quotations and the & symbol. For example, “hurricane” & “loss”.
Using predictive modeling to augment claims management systems is expanding beyond insurers and third-party administrators, those familiar with the technology say.
By helping claims professionals get a better sense of the potential severity of a claim and the probabilities of subrogation and litigation, the potential for analytics to reshape the claims process for the better is profound.
Jane Tutoki, New York-based global head of claims operations at Chartis Inc., said using predictive analytics is the key to keeping the costs of claims as low as possible.
“You can use analytics for all aspects of claims — it's not just for first notice of loss,” Ms. Tutoki said. “Whether it's subrogation or triaging of claims or recognizing fraud, analytics is the name of the game.”
“The key to using claims analytics is to make sure that you are supplementing the experience of the examiner,” said Keith M. Higdon, senior vice president of decision support services at Memphis, Tenn.-based Sedgwick Claims Management Services Inc. “You're not trying to replace it.”
The ultimate goal of analytics in the claims process is to make the best use of examiners' time and identify opportunities to achieve better outcomes, he added. “You can build the world's greatest model, but unless you have a mechanism to deploy it and an intervention that can accurately leverage that information, the model does you no good,” Mr. Higdon said.
Accordingly, when to use analytics is an important consideration. While improvements in the underlying computer processing power and systems architectures have enabled even computationally dense predictive models to be run in nearly real time, faster is not always better, Mr. Higdon said.
Because the specifics of a claim likely will vary during its life cycle, it is important that that analytics functions evolve while the data is changing, he said.
“Intake is not always the best snapshot of what a claim will eventually look like,” Mr. Higdon said. “Waiting until a more thorough claims investigation is done can sometimes improve the output.”
Timely and adaptive analytics is hugely important, said Scott Robinson, Los Angeles-based principal of analytics at CS Stars L.L.C., a Marsh Inc. unit that serves the technology needs of risk managers.
“Scoring a claim is not a one-time event,” Mr. Robinson said. “There are multiple points in the life cycle of a claim where you may be looking for different answers.”
Mr. Robinson said he sees a broader group of users who are interested in applying analytics to the claims process.
“Claims predictive analytics is fairly well established among third-party administrators and insurance companies, but what we are working on is moving that down to the individual company level,” he said.
A September report by Boston-based insurance advisory firm Strategy Meets Action found that 52% of commercial property/casualty insurers have invested in data and analytics for the claims process, second only to the 73% that have made such investments for underwriting coverage.
One potential constraint on greater use of analytics is the relative scarcity of people to formulate predictive models. While these skill sets may be prevalent at insurance companies, chief financial officers and risk managers at smaller firms may struggle to find and afford the internal talent to build such models.
“Model creation is still the realm of actuaries, data scientists and modeling professionals,” Mr. Robinson said. “If you don't apply the rigor of creating models properly, you run the risk of bad results.”
Another possible challenge for smaller entities looking to use analytics in the claims process is a lack of data. One solution is to pool anonymous data from several entities.
Mr. Robinson said CS Stars has data from 14 million claims on its software for benchmarking and building generic models that can help clients better predict the outcome of claims.
“If a client has some claims, but isn't quite over the threshold of data needed to build an effective model, we can help fill in those gaps,” Mr. Robinson said.
Likewise, Denver-based Valen Technologies Inc., a data and analytics provider for property/casualty insurers, pools data from consenting, similar clients to help to ensure that there is sufficient data for each to feed its predictive models.
Valen President and CEO Dax Craig said he expects businesses' use of analytics in the claims process to follow a similar adoption curve as in insurance underwriting, where the use of analytics has become pervasive after years of sporadic adoption.
Mr. Craig said analytics is especially promising in areas such as medical management claims fraud and workers compensation. Here, armed with knowledge of a how a medical provider's treatment patterns deviate from a statistical norm, an adjuster may better spot overbilling for a given treatment.
“When it comes to medical claims, an adjuster is often up against a physician without the depth of knowledge about the particulars of that claim,” Mr. Craig said. “Predictive analytics changes that asymmetry of information.”
The data used to populate predictive models can come from either internal or external sources, and in either structured or unstructured formats.