Workers’ Comp and Predictive Modeling

In 2002, faced with the monumental challenge of rebuilding a team with the lowest salary cap in major league baseball, Oakland A’s General Manager Billy Beane bucked conventional wisdom by employing data analytics in groundbreaking ways to uncover hidden predictors of player performance and field a winning team. Faced with their own challenges of escalating premiums and soaring medical costs, workers’ comp carriers need to be just as bold and creative.

While data analytics derived from predictive modeling is being successfully applied in many industries and lines of insurance, such as group healthcare and personal lines insurance, the practice is only slowly gaining traction in the workers’ comp industry. In many cases, high development costs, resistance to drastic change, and the leap of faith required to integrate advanced analytics into ongoing business operations are deterrents to carriers’ willingness to invest in new practices. But the rewards can be great.

In the workers’ comp field, carriers can use predictive modeling to predict the overall cost of a claim and prospectively identify claims that may incur high medical costs far earlier in the claims management process than before. Such information is critical to carriers in setting reserves, designating resources, and determining case settlement strategies. Predictive modeling can also be used to revolutionize the accuracy and quality of business functions like premium auditing and fraud detection by allowing a company to see hidden exposures before they surface and align its strategies accordingly to help manage costs.

So What Is Predictive Modeling?

Simply put, predictive modeling is the art and science of using advanced statistical analysis of past experience to find hidden connectors that more accurately predict future outcomes. Recent expansion in capabilities to collect, store, and manage vast amounts of data—and new techniques to evaluate that data—open the door for those predictions to become more precise and more powerful. Predictive modeling can more efficiently examine all available historical data assets and identify a greater number of potential predictors, take into account more interactions among predictors, and assign them with varied relative importance.

Traditionally, for each of their business functions, workers’ comp carriers have used sets of management rules and guidelines developed from experience over time. Claims, fraud, and audit managers apply these rules to separate claims and policies posing no adverse risk from those that may have a substantial impact on the business. While the traditional rules-based approach can point a carrier’s management team in the right direction, many decisions continue to be made subjectively, and the potential for the “not so obvious” exposures to slip through the cracks still exists.

Given the volume of data in workers’ comp, the regulatory reporting mandates that require carriers to stay organized, and the technology now available to analyze the data, why are decisions still being made subjectively? Resistance to change is a big factor. It’s the same resistance to change that Beane ran into in his clubhouse. Beane realized that his scouts were making decisions on player value and performance subjectively, driven by tired traditional methods, personal experience, and gut feelings — not by the vast amount of statistics right at their fingertips and not by applying science to the numbers. Guided by his Yale-educated assistant general manager, he took a leap of faith by playing a numbers game to build a new business model.

It’s time for workers’ comp carriers to do the same. There’s no other line of insurance as data-intensive as workers’ comp. Predictive modeling thrives on data and can provide carriers with a sophisticated and scientific approach to managing business operations, illuminating hidden exposures, and allowing internal teams to make more accurate, fact-based decisions. Workers’ comp and predictive modeling are a perfect match, and many functions can benefit from this sophisticated technology.

Setting Reserves, Case Management, and Settlement Strategy

The ability to predict the overall cost of a claim and prospectively identify claims that may incur high medical costs early in the claims management process is critical for carriers to effectively set reserves, designate resources, and determine settlement strategies.

Reserving, one of the most difficult aspects of claims management, is often the most important because it directly affects many aspects of the overall carrier’s health. For example, under- or over-reserving can result in unanticipated consequences, such as insufficient premium coverage, unjustifiably high rates, or an inaccurate outlook of the carrier’s overall profitability. Over time, continued reserving issues across multiple claims can amount to a much larger threat to maintaining solvency and competitiveness in the market. The key is to be as precise as possible in the reserving process to maintain an accurate projection of total liabilities and available funds at all times.

Using predictive modeling for early determination of the complexity of claims can provide immediate feedback to improve a claims manager’s ability to make smart choices for reserves. Predictive modeling can score claims that have the potential to “explode” in cost. At the same time, it can assign accurate dollar and lost-time ranges for both indemnity and medical benefits that can be expected.

For example, on the surface a claim may look like a routine back injury strain requiring minor medical treatments, but when combined with other predictors, such as the claimant’s distance from work, smoking habits, marital status, or number of dependents, the claim can evolve into a longer-term, high-cost case. Data predictors can help connect the dots between these seemingly unconnected factors and actual outcomes.

While the frequency of “exploding” claims may be low, they can often account for the majority of losses over a carrier’s entire book of business. Using a predictive model to help adjusters easily identify these types of claims early in the claims management process can allow for proactive measures in setting accurate reserves, gathering the right resources to triage claims effectively, and mitigating the extent of losses down the road.

Creating an environment where injured workers with a high likelihood of long-term disability and medical issues can receive the most appropriate treatment available is crucial to achieving the best outcome and mitigating cost.

Carriers can also use predictive modeling to analyze medical data collected during the claims management process. By linking claims to single providers by elements such as the type of diagnosis, treatment, amount charged, amount paid, and the result of the treatment, a predictive model can help determine which doctors consistently maintain the best outcomes by injury type and create a provider network of the most successful physicians for case managers to use.

Predictive modeling can also flag which cases may benefit the most from case management services. The flagging process, coupled with a reliable provider network, can help a carrier ensure that its most costly claimants are receiving the most effective treatment to recuperate, return to work as soon as possible, and avoid unnecessary future procedures.

Premium Auditing

As the 2002 baseball season went on, Beane expanded his use of data analytics to improve his team’s on-base percentage by predicting what type of pitches would be thrown, which players would have the highest probability of hitting certain pitches, and when players should swing or take pitches. He armed his team with this information so they could use it on the job.

Similarly, carriers can expand their use of predictive modeling and data analytics to arm their audit teams with measures for determining which companies are the best candidates for audits, the order and type of audits to conduct, and whether or not the effort is worth the recovered premiums.

The analyses from predictive modeling tools provide a more precise, individualized estimation of the additional premium potential that exists in each policy. Unlike conventional rules-based systems, predictive modeling can provide a far more scientific and accurate approach to choosing when and whom to audit by scoring potential candidates with a ranking and range of recovered premium and by determining the most efficient type of audit to conduct. Armed with this information, premium auditors can assist in claims triage by properly evaluating each policy for premium potential early in the audit assignment and maximizing the efficiency of their resources to manage costs. This places the premium audit team in a position to capitalize on available opportunities and benefits the carrier by uncovering hidden premium dollars faster, increasing investment income, decreasing cost of operations, and lowering loss ratios.

Fraud Detection

There’s no shortage of fraud schemes or perpetrators of those schemes in workers’ comp and the healthcare industry. New schemes pop up all the time. And for a carrier, the longer it takes to discover a fraudulent claim, the larger payouts can be. Because early detection of fraudulent and abusive claims is critical to containing costs, predictive modeling can also be highly effective in the battle against workers’ comp fraud.

Predictive modeling can identify fraudulent, abusive, and high-risk claims much earlier and with a higher degree of accuracy and quickly process the majority of claims without adjuster intervention. Predictive models look at thousands of claims variables simultaneously, recognizing complex, subtle patterns by connecting variables to other variables where there is a relationship, such as all claims associated with a particular medical provider, links between individuals that appear in several claims, or a claimant linked to several addresses or Social Security numbers. Using those connections, the models score claims indicating the likelihood of fraud. This score can arm adjusters with a quick and easy way to distinguish between suspicious and meritorious claims and help them determine what action to take, such as routing high-scoring claims to appropriate specialists in case management or special investigation. By accurately identifying high-risk claims, predictive analytics can make it efficient for insurers to process and close a vast majority of claims faster and with more confidence.

A Brave New Business Model

The workers’ comp landscape continues to challenge carriers with escalating costs and shrinking margins. Adopting—and adapting—new technology like predictive modeling into workers’ comp business functions will become essential for carriers not only to maintain a competitive edge, but to survive.

While change doesn’t come easy, carriers must not let past successes overshadow the need to look forward, think creatively, and adapt to growing challenges.

The technology exists—and the potential for using predictive analytics in the workers’ comp arena is wide open. But it will take courageous and innovative leaders to embrace this technology and inspire and guide their organizations—much the way Billy Beane did.

Many in professional baseball were convinced Beane would fall flat. He replaced the three star players he lost to higher paying teams with castoffs from other teams, and the A’s got off to a rocky start. But disruptive change takes time and perseverance. Beane learned that the only way to truly integrate the practice into the mindset of his players and coaches was to continually work with them to further their understanding of the team’s ultimate goal and what the future held for the organization.

As the season went on, the numbers started to click. The A’s started winning. The small market team without superstars won an unprecedented 20 consecutive games and finished first in the league. Other general managers had to take notice: The status quo would leave them vulnerable. There was a brave new business model in professional baseball.

Predictive modeling offers workers’ comp carriers the power to change the status quo and create their own brave new business model.

Nicholas Floeck is associate director, ISO Workers’ Compensation Solutions at Verisk Analytics.