Big data is bringing big benefits to organizations that dig in deeply to analyze oft-hidden patterns and other useful information buried in their claims data, particularly highly predictable workers compensation exposures. The companies are discovering unknown correlations that guide more informed risk management processes to reduce claims frequency and severity.

In some cases, these evaluations are performed on behalf of corporate risk managers by their insurance brokers, which have upgraded their Risk Management Information Systems (RMIS) tools to afford big data analytics. A voluminous amount of unstructured, semi-structured and structured data is fed into such systems to discern unusual claims activity or repeat scenarios, which then directs actionable tactics to mitigate and reduce the underlying exposures.

“In the insurance space, it is in the claims area where there is probably the greatest volume of data for loss prevention purposes,” said Claude Yoder, head of global analytics at insurance broker Marsh.

For now, this analytical activity is confined largely to casualty claims, such as general liability and workers compensation. Brokers say the service is assisting their more sophisticated risk manager clients to attain competitive advantage over peer rivals, reducing claims frequency and severity to pare insurance costs, and increase worker safety, satisfaction and productivity—with bottom line benefits.

“For large organizations with lots of loss information coming from many sources, we can come in, collect and assimilate all this loss data to help their risk managers track, manage and control their claims,” said Marc Hindman, chief risk control officer, in the risk control and claims advocacy practice at insurance broker Willis in New York. “This, in turn, helps them achieve better outcomes, improving insurance treatment.”

Gold in them Hills

Many companies today may be surprised to learn they have at their behest reams of claims data that can be turned into insightful information for decision-making purposes around worker safety. The problem is that much of this data is buried in spreadsheets or concealed within organizational departments that then fail to share the facts and figures.

“Unearthing this data in a consistent fashion for analysis purposes can yield surprising results,” Yoder said. “For instance, you can create a loss severity file using predictive modeling that can alert you when certain factors are arising that may result in a financially severe claim.”

A case in point is a company that over many years has experienced a large volume of a particular type of claim, such as workers who have experienced back injuries on the job involving the same vertebrae and the same prescriptive narcotics. Such signals may indicate a future severe claim. Armed with this information, the risk manager can determine causal factors for the injury, and put in place ergonomic programs that reduce the incidence of this type of injury in future.

“You create a severity model using the data analytics that allows you to predict as early as possible in the life of a claim those that will become this particular type of back fusion cases,” Yoder said. “By taking advantage of the information you have across different points in time, you can then triage [the claim] the right way. At the `first notice of loss,' you now have a crystal ball that tells you `here is a certain kind of back fusion case, so you better put it in the hands of your most experienced adjusters, nurse case managers, legal counselors, and so on.”

The goal is to work backwards from historical claims to discern the factors leading to each claim's occurrence. In this activity, both structured data like actual numbers, and unstructured data like a claim examiner's notes, play significant roles. “There is an incredible amount of information embedded in the claim examiner notes—not just the `first notice of loss' data, but the medical information coming in over time on a case-by-case basis,” Yoder said. “By quickly analyzing this data, risk managers can glean proactive measures to reduce the associated (insurance and business disruption) expenses.”

Paralysis by Analysis

Unfortunately, while organizations have the data, they often lack the “crystal ball.” Many companies attempt to collect and assimilate vast streams of loss information from multiple sources in a relational database to recognize relationships between the data elements, or try to manage it “old school” with spreadsheets, according to Hindman. In both cases, they are frequently unable to distill useful information.

“It becomes overwhelming,” Hindman said. “You want to be able to get to the point where you know that 25 individuals in the Midwest division of your organization comprise 80 percent of all your `slip and fall' losses in the U.S. You can't do that with manual spreadsheets.”

There is another reason why it makes smart sense for risk managers to rely on their insurance brokers or other third party specialists to leverage Big Data for smaller and fewer casualty claims—the pronounced dearth of available IT software engineers, developers and programmers to access and evaluate multiple sources of data for analytical purposes.

Although the economy is creating 120,000 new jobs each year that require science, technology, engineering, and mathematical capabilities, US colleges are only producing 40,000 graduates with such skills annually, a shortage “that is only going to get worse,” Microsoft Chief Counsel Brad Smith told the Brookings Institution in September, 2012. Having a broker with these skills already in place, particularly for smaller and midsize organizations with strapped capital resources, means not having to hire the talent internally.

Down the line, Hindman projects that the RMIS systems provided by brokers will become more granular, presenting risk managers with increasing agility in predicting the future. “The goal is to crystal ball what hasn't happened yet but may happen in the future.”

Special Reports /

Big Data is a Big Deal

Related Stories

Recommended For You