Predictive models spot problems early
November 6, 2011 – 6:00am
A new generation of predictive modeling tools is rolling out in the workers compensation market, which experts say could make claims adjustment and management more of a science.
Modeling has been available in workers comp for years, but insiders say newer versions are more accurate in identifying problematic indemnity claims before their losses spiral out of control.
“I think everyone is always looking for that Holy Grail of information,” said Paul Braun, managing director of casualty claims for Aon Global Risk Consulting in Los Angeles.
Liberty Mutual Group Inc. launched a new predictive model this year that uses up-to-date data to calculate whether comorbid health conditions and psychosocial issues—such as obesity, depression or job dissatisfaction—could hinder an injured employee’s return to work.
Though a model Liberty Mutual introduced in 2006 considered such data, its revised VantageComp model is better at identifying claims that start small and grow slowly into larger indemnity losses, said George Neale, executive vp and general claims manager in Boston. Claim adjusters use the new data to point workers to resources that can help them recover faster, he said
“Slow-emerging claims are the ones that are a challenge for us in workers compensation,” Mr. Neale said. “And if you can understand those earlier, you can do something about them.”
He said the new model will be successful if it helps Liberty save at least 5%, or $100 million, on the $5 billion in workers comp claims it pays annually.
More insurers are using predictive models to help stem costs for workers comp claims, according to a survey earlier this year by New York-based consulting firm Towers Watson & Co. Of more than 100 insurance executives participating in the survey, 32% said they use predictive models in workers comp, up from 18% in 2009.
Aiming for accuracy
Companies also are continuing to tweak the tools to improve accuracy.
Aon’s Mr. Braun said companies are working to find data that pinpoints why some claimants take longer to recover than others with similar injuries and are evaluating whether previous data collection methods were accurate. Successful models use that information to actively help patients get well, he said.
Aon’s Early Claim Intervention model has shifted in the past two years to help adjusters better identify claims that will result in larger-than-expected losses. Mr. Braun said the brokerage recently analyzed seemingly simple medical-only claims that became complicated indemnity losses, and uses that data to spot other claims that could follow the same troubled trajectory.
Claims flagged through the process are sent to Aon’s most experienced adjusters, who then connect claimants with health care providers that can speed their recovery.
“You get them to the right doctor and get them to the right treatment instead of letting them linger,” said Mr. Braun, who estimates the model has helped reduce workers comp costs by 15% for its clients.
Atlanta-based third-party administrator Broadspire Services Inc. adapted its E-Triage model six months ago to help claims adjusters better determine whether smoking, obesity and other factors slow worker recovery times.
Gary Anderberg, Broadspire’s Philadelphia-based practice leader for analytics and outcomes, said social issues also play an important role in Broadspire’s newer mathematical formula. For instance, the more family members a claimant typically has, the less likely he or she is to return to work.
“If there are preschool-age children in your home, the fact that you’re home on a workers compensation claim may mean you don’t need to pay for babysitting,” Mr. Anderberg said. “And that can be a nice thing.”
He said Broadspire’s model, launched five years ago, is updated continually with new data. A version being developed by the TPA will allow its adjusters to predict how many days an injured worker will stay off the job and use that guideline to help prevent excessively long work absences.
Reed Group Ltd., a disability case management services firm in Westminster, Colo., released a new version of its MDGuidelines predictive model for workers comp cases in February.
Its model, used by various insurers and employers, predicts the recovery time for claims based on factors such as geography, comorbid conditions and job satisfaction. Reed Group then suggests medical resources that can hasten recovery times.
Dr. John Seymour, president of guidelines for Reed Group, said the model can help reduce a claimant’s time off work from an average of 55 days to about 30.
“We’re always striving to get that employee back to work as soon as possible,” Dr. Seymour said.
Companies expect predictive models to keep evolving as insurers and administrators discover new ways to calculate workers comp risk.
Broadspire’s Mr. Anderberg said he believes competition among companies will drive significant advances in predictive modeling in the next several years.
“There are a lot of good people out here, and some really creative thinking is going into it,” Mr. Anderberg said.