Help

BI’s Article search uses Boolean search capabilities. If you are not familiar with these principles, here are some quick tips.

To search specifically for more than one word, put the search term in quotation marks. For example, “workers compensation”. This will limit your search to that combination of words.

To search for a combination of terms, use quotations and the & symbol. For example, “hurricane” & “loss”.

Login Register Subscribe

The pros and cons of predictive modeling

Training essential to properly interpret data, speakers say

Reprints

The challenges and benefits of predictive modeling were the focus of one presentation during Business Insurance's 2012 Worker's Compensation Virtual Conference on Oct. 25.

“We want to predict the future so we can change it,” said Gary Anderberg, practice leader for analytics and outcomes at Broadspire Services Inc. “If it looks like a claim is going to have a high dollar potential, we want to do something about that.”

Mr. Anderberg, who is responsible for the company's proprietary predictive modeling system, e-Triage, spoke about how a system can help adjusters not only identify which incoming claims need additional attention and resources, but also the best way to intervene.

His goal is to make the process as specific as possible by looking at age, gender and jurisdiction, but much of this information gets lost in the information gathering process.

“So many important aspects of what make the claim perform the way it performs in a typical claims system are entombed in narrative or text boxes,” Mr. Anderberg said. “I refer to text boxes as where valuable data goes to die.”

The two other speakers for this webinar were Michael Gavin, chief strategy officer at Prium; and Carol Ungaretti, director of risk control, claims and engineering at Aon Global Risk Consulting.

Both speakers laid out some of the continued drawbacks with predictive modeling.

%%BREAK%%

“A lot of the predictive modeling we see today is highly specific and, as a result of that, it is precisely inaccurate — and that's OK as long as the claims handlers are properly trained in terms of the context and the interpretation of the output of predictive modeling,” Mr. Gavin said.

He said he had found that many adjusters rely too heavily on these tools, while others do not trust them enough.

Ms. Ungaretti brought up the point again, encouraging adjusters to use the data in the context of their own companies. Her suggestion is that adjusters evaluate their own risk characteristics and build their solutions around their own needs.

“Both an organization and an adjuster — and the carrier and TPA tied to that adjuster — really need to look at what the correlation is between, perhaps, an organization and its idiosyncrasies and those predictors,” Ms. Ungaretti said.

Later, in a question-and-answer session moderated by Sheena Harrison, associate editor at Business Insurance, all three panelists agreed on one more drawback: determining return on investment for the predictive modeling systems themselves. While Mr. Anderberg said he has found that Broadspire's E-Triage system shows about 10% to 15% cost differential, he said it's difficult to put a return on a diagnostic tool.

Ms. Ungaretti compared it with trying to implement a health or wellness program at companies.

You can view the event online at BusinessInsurance.com/CompCosts.