The economic fallout from the recent global pandemic combined with an already hardening insurance market is squeezing profit margins in nearly all industries. For claims departments in any insurance or self-insured company, predictive modeling is a fast, accessible, and effective approach to cut costs and protect against volatility—often reducing claims costs by 10% or more. But only 30% of claims professionals rely on any form of predictive analytics.¹ For the other 70% of the market, great opportunity exists to reduce costs. Even those companies that use predictive analytics may not be using the technology to its fullest potential. Innovative approaches to modeling using formerly inaccessible data can drastically improve results and drive down claims and related defense costs.

These new insurtech products overcome one of the thorniest challenges in modeling claims—inaccessible data in text form (a form of unstructured data), such as adjusters’ notes, legal invoices, and other nonnumeric information not captured in fields. Until now, claims analytics has been limited to structured data entered by staff in dropdown menus, which often falls prey to inaccurate, inconsistent, or incomplete coding. Claims descriptions and adjuster notes, for example, are key to accurate identification of high-risk, high-cost claims. Gleaning information like an upcoming surgery or attorney involvement are red flags early in the life of a claim that can be identified in claims notes and monitored or mitigated.

A similar challenge exists with defense cost management, which for years has relied on standardized Uniform Task-Based Management System (UTBMS)² codes, to classify legal expenses into predetermined task, activity, and expense codes. Like other code sets, the UTBMS codes rely on the judgment or decision-making of humans, in this case, perhaps dozens of different billing specialists who could interpret identical line items on an invoice in different ways. Computer algorithms, on the other hand, more consistently and with more granularity categorize claims defense activities into meaningful events that can be monitored to better manage the claims defense and settlement process.

Reducing indemnity costs

For claims professionals looking to reduce indemnity costs, one such insurtech product that employs predictive analytics is Milliman’s Nodal. Developed for early claims intervention, Nodal predicts future claims outcomes shortly after a claim is reported. Nodal’s proprietary text-mining algorithm extracts data from adjusters’ notes, claims descriptions, and any other text data, searching for information like comorbidities and planned medical procedures—information that might signal a high-risk claim. By culling through nontraditional data in claims files, Nodal is able to effectively develop a new source of claims intelligence that, combined with machine learning, predicts future claims outcomes faster and more accurately. These tagged claims help a company’s claims managers intervene early—either by resolving the claim, or triaging the claim to an experienced adjuster. In one recent example, Nodal identified up to 95% of an insurer’s high-cost claims and produced savings of up to 10% of an insurer’s claims costs.

Reducing defense costs

For organizations looking to manage legal defense costs, Milliman’s Datalytics-Defense is an insurtech product focused on identifying claims cost drivers but from a defense cost perspective. Its data-mining algorithms are able to extract the intelligence that was formerly locked inside defense attorney invoices, including details on depositions, motion activity, and court conferences, among other detailed claims defense events. Paired with Milliman’s subject matter expertise, this new approach brings greater insights into an insurer’s defense strategies to better manage its total spend, both defense costs as well as indemnity costs.

For example, the Datalytics predictive model incorporates certain risk-adjusted metrics in order to measure the efficiency of a company’s law firm. From the model, an expectation is developed, which can be compared to a law firm’s actual performance as a way to score and rank law firms. This data-driven performance scoring can then be used to more effectively allocate defense dollars.

Like Nodal, Milliman’s defense cost predictive model uses a sophisticated data-mining algorithm, which in this case, reads the detailed line item descriptions from defense counsel invoices. This process occurs across every single line item for every single law firm in every single state in exactly the same way, thus resulting in a uniquely consistent data set. As the algorithm reads invoices that have previously been electronically uploaded from a law firm via a web-based portal, the model extracts detailed data allowing companies to track defense activities and costs in a way that they haven’t otherwise been able to before. In this way, Datalytics not only brings much more consistency and accuracy to the data by circumventing human error but also builds new intelligence that can inform decision-making.

In their ability to comb unstructured text and transform it into new claims intelligence, Datalytics and Nodal individually go beyond any modeling approaches in the market today. Together, they can give insurers and self-insurers a compelling competitive advantage, especially in these uncertain times.

¹Algire, D.Z. 2019 Workers' Compensation Benchmarking Study. Rising Medical Solutions. Retrieved May 10, 2020, from https://www.risingms.com/wp-content/uploads/2019/12/2019WorkCompBenchmarkStudy_Rising.pdf.

² The UTBMS codeset was developed by the American Bar Association in the mid-1990s.