Prescriptive modeling adds new dimension

While predictive modeling helps claims managers understand where a workers compensation claim could end up, experts say prescriptive modeling holds the key to a question payers must consider before a claim goes south: what to do about these risks?

The field of prescriptive analytics and modeling is being used in workers comp to move beyond using data to predict what will happen to a claim to anticipate what will happen to a claim if certain interventions take place.

“(Prescriptive modeling) is taking it beyond what the predictive model defines… The prescriptive side is suggesting what you should do,” said George Furlong, Tampa-based senior vice president for managed care programs, outcomes and analytics for third-party administrator Sedgwick Claims Management Services Inc. “We are moving down the path from knowing that there is a risk to identifying what to do differently about that risk.”

“We want to help deploy strategies at the desk level to understand what we can do differently with this claim and to help educate the adjuster and nurses,” he said on the relatively new practice of prescriptive modeling.

Prescriptive modeling has its roots in predictive modeling, which payers have been engaged in over the past decade (see related story), experts say. But a surge in data collection in workers compensation, along with technological advances in predictive modeling for injury claims, is allowing payers to better understand where a claim could go wrong and why.

“It’s growing leaps and bounds,” said Jayant Lakshmikanthan, Santa Clara, California-based founder and chief executive officer of Clara Analytics Inc., which partners with Aon Inpoint Claims, a data and analytics arm of Aon P.L.C., on improving workers compensation outcomes by acting as an “air-traffic controller” on claims coming in and what to do.

“I think what we have seen in the last few years is the increase in the awareness of being able to use this information,” he said.

Predictive modeling alone is similar to the process that tells hurricane forecasters, using mounds of data and the paths of previous similar storms, where a certain hurricane is likely to hit — an analogy Jeffrey White, Rolling Meadows, Illinois-based senior vice president and product manager for workers compensation at Gallagher Bassett Services Inc., likes to use when he talks about the trend in managing care and outcomes for injured workers.

“The spaghetti models (for hurricanes) come out when the storm shows up, (and) it’s not always right but you kind of know you can start making preparations” depending on where you live, he said. “That it’s coming down the path, I know I have to get ahead of it … (With workers comp), what you are trying to do is take historical data and learn trends so that you can better understand what the possible outcomes will be.”

From there, payers must deploy expertise to heed the warnings and act on the claims that could pose problems, but predictive modeling fails to address what comes next, experts say.

Predictive modeling is not the tool to keep a claim from getting complicated, said Mark Moitoso, Atlanta-based executive vice president and risk practices leader at Lockton Cos. L.L.C. Experienced claims adjusters and managers are at the heart of why predictive modeling works to close claims, he said.

From there, payers must deploy expertise to heed the warnings and act on the claims that could pose problems.

“(Predictive modeling) does not take away the need for a (claims professional) to figure out what the next steps are to avoid the identifying factor (in increasing costs),” added Paul Primavera, Washington-based executive vice president and national risk control services group leader at Lockton.

Melissa Dunn, Chicago-based regional claims advocacy leader in the Midwest/ Great Lakes region for Arthur J. Gallagher & Co., who spent parts of her 28-year career in the risk management and workers compensation departments for employers such as Ford Motor Co. and McDonalds Corp., illustrated a typical example of taking steps beyond predictive modeling to prescriptive.

“You have a 55-year-old in Pennsylvania with a back injury after falling from a ladder; you might want to get a nurse involved and you might want to proceed to surgery quickly because nine out of 10 times, these cases wind up in surgery after months of aggravating the injury,” she said. “Predictive analytics doesn’t save money; people do. It isn’t the data alone; it’s responding to that data, and the adjuster taking action.”