By Peter Rousmaniere, an expert on the workers' compensation industry.
Predictive analytics for workers' compensation claims assess the chances that a work injury will depart from an average expected profile as measured in duration of disability, medical utilization, litigation or other outcomes.
Managers can focus on high-risk claims that deserve special attention.
To be effective, the prediction system can't be stumped by the innumerable unique features of each claim, and the claims payer must know what resources to deploy quickly in response to high-risk claims.
The observations of the largest workers comp carrier, a supermarket chain and a regional insurer confirm that predictive analytics is rooting deeply into the claims payer community.
Liberty Mutual, which wrote about $4 billion in workers' compensation insurance in 2010, said it has been working on a predictive system since 2004.
George Neale, executive vice president and general claims manager of Liberty Mutual Commercial Markets, said that modeling must be a key part of an integrated claims-management approach.
Modeling "enables us to close workers' compensation claims faster and to produce lower average paid costs per claim than our competitors," he said.
This approach allows the insurer to apply the right resources to each claim at the right time to better manage outcomes and costs. The model is seamlessly integrated with Liberty Mutual's claims-management system.
The insurer's huge archival database -- it evaluated more than 825,000 lost-time claims and 140 million medical-billing transactions -- allowed it to find highly predictive variables and to understand their interplay to spot claims likely to escalate in cost.
"Today," Neale said, "we can identify claims with the potential to be high cost as early as claim intake. In addition to intake, the model reviews each claim at four other specific points in its lifecycle."
Safeway, the largely West Coast supermarket chain, uses a radically different approach. It asks occupational medicine clinics to prepare a specially designed profile of Safeway employees with fresh injuries. Bill Zachry, Safeway's risk manager, said that Safeway first introduced the system in 2008 in cooperation with some of Kaiser's occupational medicine clinics.
"What matters is the injured workers' underlying inability to cope," Zachry said. The profile was originally given only to employees with back injuries. Now anyone with a musculoskeletal problem takes the profile, called STarT, which asks 11 questions.
He recently analyzed 77 cases that the system identified as high risk. Kaiser had provided more physical therapy and psycho-social counseling, and store supervisors were more attentive to the workers.
All of the 77 workers returned to work. Only two of them had surgery, a popular -- and questionable -- treatment in California.
A.I.M. Mutual, a Massachusetts-based monoline insurer, has been partnering with Best Doctors Occupational Health Institute for several years to develop predictive tools and other managed- care programs.
Michael J. Shor, managing director of the Institute, said that an early claims assessment, introduced in 2005, "may include third-party data, very early drug utilization, psycho-social factors, co-morbidities, primary and secondary diagnoses and nature of initial provider. The assessment might be done by a claims adjuster, medical case manager or initial medical provider.
"Specific bio-pyscho-social risk factors dramatically increase the risk for therapeutic failure regardless of diagnosis or treatment," Shor said. For high-risk cases, the insurer deploys a response to address these factors.
Shor has some suggestions for other claims payers. He said that predictive modeling needs to be hard wired into the claims-management procedures. Claim executives need to be involved in its development and introduction and see it as skill-enhancing. The tool must be very easy to use and provide immediate feedback. Data should be able to be aggregated by employer group, and industry segment. And it should not be over engineered, he said.
As predictive analytics becomes the norm, look for greater subtlety in both prediction and in responses to red flags.
September 27, 2011
Copyright 2011© LRP Publications