(Image credit: Shutterstock.)
Sometimes an injury is not what it seems to be—or can become something more than it originally was. In a workers’ compensation context, moderate injuries can turn into large-scale losses owing to a combination of factors. Midwest Employers Casualty (MEC, St. Louis), a Berkley (Greenwich, Conn.) company, created the XCEL Analytics platform to identify such cases in advance, and the provider recently added artificial intelligence capability to enhance the platform’s performance. By identifying the types of claims that can get out of control, MEC helps its clients and their injured workers minimize cost and achieve better outcomes through early intervention.
The origins of XCEL Analytics go back to 2011, relates Brian Billings, VP, Predictive Analytics, MEC, who built the platform’s first predictive models. “As an excess workers’ compensation carrier dealing with very large loss claims, we sought a way to use existing data to identify those cases in advance,” he says.
MEC serves individual and group self-insured clients that typically have self-insured retentions (SIRs) starting at approximately $500,000 and up—which puts them on the hook for large losses. Some of the most severe injuries, such as brain trauma, are typically reported very quickly, according to Billings. By contrast, the typical case that MEC XCEL Analytics works to identify is a musculoskeletal injury that may start small.
“It could be an injured worker with a back injury who receives conservative treatment that doesn’t work, then there’s surgery, and maybe eight years later, at the end of which it’s a very large loss,” Billings explains. “The thinking is that if we can identify such cases within six months, we can get the injured worker effective treatment which means a possible better outcome for them and their employer.”
Building a Team
Billings created the first-generation models in 2011 and then secured approval to hire a PhD for the next generation. “After bringing in the heavier guns to build out the models, we did work to prove them out,” he relates. “In 2013, we got approval to build out a team for predictive analytics, including a credentialed actuary and were charged with automating the platform.”
The system went live in June 2013 for MEC’s individual and group self-insured clients and TPAs. In 2016, MEC rebuilt XCEL Analytics predictive models, updated the platform’s data infrastructure and moved it onto the company’s new data platform. In 2018, MEC added severity ranking and alerts about factors that can affect a claim’s propensity for higher loss. As the first predictive analytics platform offered by an excess workers’ compensation carrier, XCEL Analytics has proven a market differentiator for the provider, according to Billings.
“It helps those clients with large volumes of claims who need be able to analyze open claims inventory to identify those with the potential for large losses,” he says. “We deal with large complex, claims day in and day out, and we let our clients and their TPAs know that we can help them get involved to bring about better outcomes.”
In early 2020, MEC incorporated artificial intelligence (AI) into its severity ranking and added nearly a dozen new alerts that users can view in a personalized dashboard. “We integrated adjuster notes as an additional input to the predictive models, and we’ve integrated the output of those adjuster predictive models into that ranking process,” notes Billings.
The alerts work as a two-step process, identifying the claims and the specific issues within them that warrant intervention. MEC XCEL Analytics currently has over 20 alert categories to flag for intervention. “We have partnered with our Chief Medical Officer, Dr. Fernando Branco, and other medical and claims staff to identify factors,” Billings says. “For example, extended opioid use alerts went live in Sept. 2018, and we’ve continued to enhance that and others by identifying comorbidities such as hypertension and diabetes, as well as complications such as depression and PTSD.”
The alerts give claims adjusters the opportunity to speed up the intervention process. “It tells the adjuster why the model is picking up on the claim, and what can be done about it,” Billings explains. “The more our clients know, and the earlier they know it, the better they can manage each claim to the best possible outcome.”