In the face of increasing costs, new predictive modeling tools allow carriers to realize more efficient claims handling and lower loss ratios.
“Strategy, without information upon which it can rely, is helpless.”— Colonel William Donovan, U.S. war hero, intelligence officer, diplomat, and head of the Office of Strategic Services during World War II
Colonel Donovan, the “father of American intelligence,” understood that reliable information is the foundation for military success. As claims executives battle for advantages to improve their bottom lines, they are turning to more sophisticated ways to mine data, better inform their decision-making, and positively affect their claims-handling processes.
Writers of workers’ compensation policies have been waging a war against spiraling claims costs for years. Many organizations, however, are not armed with sufficient data or the supporting systems to identify and proactively address potentially burdensome claims. Many of those claims initially even appear innocent but then, over time, spiral to become complex and costly.
Some organizations have embarked on the use of predictive analytics in claims and, if they have sufficient data, have seen positive results. However, given ongoing market challenges and continually emerging technologies, even those companies are left to wonder whether that is now good enough.
According to the National Council on Compensation Insurance (NCCI), the increase in medical and indemnity costs in 2012 was a modest three percent and one percent, respectively. The staggering year-over-year increases of 5.7 percent for medical and 3.2 percent for indemnity from 2002 to 2011, coupled with questions surrounding regulations taking effect as part of the Patient Prorection and Affordable Care Act, suggest that there is additional opportunity for improvement. With the combined ratio still high at 109 percent in an environment of continued low interest rates, there is very little room for error. Using predictive analytics for immediate identification of claims that might ultimately prove difficult is a key to improving results.
The only way to prevent certain cases from spiraling out of control is early intervention and proactive claims handling. Given that roughly 20 percent of claims are driving an estimated 80 percent of losses, insurers that quickly and correctly identify which claims are likely to prove complex and costly can manage them for the best outcomes possible. The key to early identification is having sufficient and reliable data to predict and classify with accuracy the cases that may prove costly.
Claims may escalate due to many factors, including: unnecessary treatment; obesity; psychological issues; social circumstances; comorbidities; excessive use of medications; secondary gain; and fraud. Critical data points may not be easily available, while some data may be hidden in the first report of injury, medical records, PBM reports, bill review records, and so on.
As predictive modeling practices evolve and new tools are introduced, even carriers with less data can take advantage of the benefits of more sophisticated techniques. According to Karlyn Carnahan, former principal with Novarica, fewer than 15 percent of workers’ compensation carriers currently utilize predictive analytics in their claims processes, but many are now beginning to embrace modeling tools. Additionally, she notes that carriers applying such techniques “are reaping huge rewards and generating significantly lower loss ratios.”
Tools now available can improve on “good” results by leveraging internal and external data to provide greater insight into claims and the factors that may lead to escalating costs.
Good Information Equals Good Action
With large case loads and numerous fires to put out each day, adjusters understandably struggle to spot the potential complexity of every claim at the outset. That’s why predictive analytics is essential—it helps ensure complicated claims don’t fall through the cracks. That can be achieved only if analytics is embedded into the claims organization’s workflows and processes.
Any effective predictive analytics program must be able to meet three essential criteria:
Many traditional tools will meet the first two criteria by modeling their clients’ own data using multivariate analyses and providing a scoring output. Some may even combine the concept of machine learning—including algorithms for checking, validating, and testing data—to identify risk characteristics that relate to profitability.
While the traditional approach may have yielded some successful results, a carrier’s data limitations may produce less reliable outcomes. That’s where new models are emerging to meet all three essential criteria.
If a carrier has access to a more reliable representation of industry experience based on a larger array of data, those predictive scores offer more value and allow claims professionals to get cases into the right channels sooner.
Here are some suggestions that may be useful in developing or enhancing a predictive analytics implementation.
First, systems integration should be seamless. Output produced should make the claims department more efficient and effective. Scores should be simple enough to allow claims administrators to make quick, logical sense of the results and take decisive action in triaging and managing cases. The claims organization should see significant and measurable impact on loss results by aligning the appropriate resources to the right claim at the outset.
Second, if you elect to use a vendor, the following suggestions also should be considered.
Given all of Colonel Donovan’s military successes, consider what he might have been able to do with today’s analytical tools and technologies to better understand the battlefield, place the best people in the right positions, and be more proactive on a day-to-day basis.
Not coincidentally, all of those same pursuits can enhance the claims handling world, as well. Granted, claims tools may not decide the fate of the free world, but they can successfully help turn the tide against spiraling claims costs—and maybe even one day win that war, too.
