Data-rich compensation items for workers are a fruitful field for analysis that can target claimants and others who are best placed to deal with them, help identify qualified physicians and lawyers, and provide other insights.
The results for insurers and employers are healthier workers and lower costs, among other benefits, say sources.
Predictive analytics helps adjusters make decisions, says Virna Rhodes, Marlton, New Jersey-based senior vice president, occupational injury claims for Liberty Mutual Insurance Co. and its Boston-based third-party administrator Helmsman Management Services LLC.
“Data analytics are only as good as the people who use them,” she said. If used effectively, they can help “get the claim in the hands of the right expert as soon as possible.”
;Rhodes said Liberty Mutual’s claims system is “only flooded with notes and data” that artificial intelligence analyzes to flag claims based on potential severity, the likelihood of an injured employee returning to work, liability for a claim and whether subrogation is probable or fraudulent. occur.
Analytics can generate significant savings, sources say.
“You’re trying to claim the right person at the right time,” said Adam Wesson, head of claims management at Verisk Analytics Inc.’s North Reading, Massachusetts-based ISO Claims Partners. “The sooner you involve the right resources in handling the claim, the better results you are likely to get.”
Skip Brechtel, vice president and chief information officer at Cannon Cochran Management Services Inc. in Metairie, Louisiana, said an analysis of Cannon several years ago showed that claims with two or more co-morbidity factors cost as much as seven times more than those without them. . In 2014, CCMSI began using an algorithm that took into account 70 co-morbidity data fields to help determine if claims for damages could become serious and require closer attention from management.
“And that’s how we got started this way,” which led CCMSI three years ago to send around 400 claims data items and other information daily to Boston-based analytics provider Gradient AI Corp., where artificial intelligence identifies drivers of claims costs, predicts . results and provides guidance for reserving.
“As we all know, artificial intelligence in the use of predictive analytics has come a long way in the last four to five years,” Brechtel said.
Stan Smith, CEO of Gradient AI, said that its insurers and TPA customers typically shave off a 3% to 5% discount on claims costs, and some see savings of as much as 14%. “It has been very strong in any case, but it has been very significant in some cases,” he said.
“These models get smarter the more data you give them,” said Joe Powell, senior vice president of data and analytics at Gallagher Bassett Services Inc. in Fort Wayne, Indiana. TPA feeds its machine learning model with medical billing, diagnoses and other injury-related information to create tens of millions of records that can be used to study results.
A valuable use of analysis is to help identify physicians and attorneys who contribute to favorable outcomes for employers, said Mr. Powell.
Gallagher Bassett selects its medical network providers based on performance and quality data, he said.
Suppliers are evaluated and rated on how carefully they follow medically based guidelines when treating injured workers, said Mr. Powell. “And then we can rank and rate our suppliers according to how often they process according to the book,” he said.
“It’s very similar on the part of the law firm,” said Mr. Powell. “We look at the issues they handle and the results they pursue and we measure their performance.”
For both physicians and attorneys, “we adjust these results based on the complexity of the cases they handle,” said Mr. Powell. An algorithm ensures that doctors and lawyers who handle mostly complex cases are not necessarily ranked lower than those with simpler and easier cases to resolve, he said.
Analytics can also help manage the risk of workplace injuries, Wesson said. The models provide “a wealth of information” that employers can use to identify the type, frequency and severity of injuries among their workers, he said. By revealing these trends, risk managers can better understand how to prevent injuries.
The structure of the models varies.
“The most primitive approach may be to set up a handful of business rules and call them predictive models,” an approach that may have some value, “said Joel Raedeke, Chicago-based technology and computer science director at Crawford & Co. TPA Broadspire Services Inc. “It’s a kind of window painting I’ve seen.”
A more robust approach is to build a multivariate predictive model that is powerful and, if maintained properly, can integrate new features as they evolve over time, Raedeke said.
Such models can be expensive and labor-intensive if they do not use artificial intelligence to keep them current, he said.
Source link
