Case Study: Efficiency and control through Milliman Mind
This briefing note provides a short case study on migrating an Excel/VBA model to Milliman Mind.
Over the past several years, workers’ compensation insurers have been on a winning streak. Combined ratios are down, profits are up, and frequency continues to decline. As a result, premiums have continued to fall over the past seven years, with the National Council on Compensation Insurance (NCCI) dropping rates 7.2% from 2019 to 2020. To maintain recent underwriting performance in the face of these rate cuts, insurers must innovate to reduce claim costs. A recent A.M. Best article noted that “the highest performing workers’ compensation insurers are effectively using data analytics and predictive modeling to manage claims.”1 Claim analytics using artificial intelligence represents a proven method to reduce claim severity and improve the efficiency of claim-handling resources.
Workers’ compensation results have been a bright spot for insurers compared to other lines of business, which for the most part have weighed down overall profitability. Since 2015, the industry calendar year combined ratio for private workers’ compensation carriers has been well below 100%. In 2019, those carriers had an 85% combined ratio—the third year in a row under 90%. To put this in perspective, the 2018 and 2019 results are the lowest combined ratios since the 1930s. These results reflect continued decreases in claim frequency due to improved worker safety and claim severity trends that are moderate by historical standards.
However, the workers’ compensation market may experience headwinds in the near future, as COVID-19 injects uncertainty in an otherwise profitable market. Ultimate costs for COVID-19 claims are unclear due to changing infection rates, unknown long-term effects, and coverage issues. Non-COVID-19 workers’ compensation claims are also affected by the pandemic, mainly due to deferred medical care, which could lead to longer cycle times, higher rates of lost time, and changes in utilization.
A low interest rate environment also puts pressure on profitability, as workers’ compensation is a long-tailed line and heavily reliant upon investment income. Premium volume is also expected to decrease due to higher unemployment and decreases in business activity and hours worked.
The unpredictability resulting from the pandemic and potentially shrinking margins due to lower premium rates place increased pressure on insurers to optimize controllable aspects of their operations, such as claims.
Over the years, insurers have made some progress in managing claim severity with manual techniques like nurse case management and return-to-work programs. These methods, though, are largely reactive rather than proactive. Predictive analytics can proactively identify a high-cost claim shortly after it is reported, and well before large costs have been incurred. This provides an opportunity to manage these costs to the insurer’s benefit.
Historically, rich sources of data in adjusters’ notes and other textual documents have often gone overlooked. Because of time and expense constraints related to reviewing extensive text, key cost triggers have remained buried in claim files and, thus, unavailable to claim managers.
This obstacle can be effectively eliminated with the use of predictive models that now make it possible to tap into this valuable unstructured data. Using machine learning and natural language processing, a branch of artificial intelligence, predictive models can now read through formerly underutilized data such as adjusters’ notes and other text sources to identify significant cost drivers like comorbidities, opioid use, and major medical procedures that may signal high-cost claims.
Identifying claims with potentially high costs can give insurers incredible insight, because between 5% and 10% of claims account for 80% of loss and allocated loss adjustment expenses in workers’ compensation.
In the past, efforts to identify these high-cost claims have been stymied by inaccurate or inadequate information in predefined fields within claim files. In many cases, coding limitations have left claim managers in the dark as to what body parts are injured in a “multiple body part” claim. For example, a finger/hand injury has an extremely different cost profile than a shoulder/neck injury. Being able to identify these cost drivers is crucial in claim cost management. Timing is also important. Not only can predictive modeling identify key cost triggers, it can do so long before a transaction has even been recorded in the claim file. This strategy gives adjusters a jump on discussions about the need for and appropriateness of the treatment.
In scanning these formerly hard-to-access data sources, predictive models effectively create a new source of intelligence. The results can be used to better inform claim managers’ decision-making in the allocation of resources and the appropriate use of more traditional claim management techniques, which can be brought in much earlier in the claim process when they are most effective.
A proven advance in claim management, effective predictive models have been shown to substantially reduce claim severity. In these uncertain times, managing claim costs is perhaps one of the most effective ways to stay vigilant, remain competitive, and prepare for the future.
1Blades, D. Lentz, J. Mangano, D. Declining rates could threaten profitability of workers’ compensation line. Best’s Market Segment Report. (December 20, 2019). Retrieved on August 18, 2020, from http://www3.ambest.com/bestweekpdfs/sr536849119015full.pdf.
WC Market Outlook: After years of successive rate decreases, how can workers’ compensation insurers stay ahead of the curve to maintain profitability?
Claim analytics using artificial intelligence represents a proven method to reduce claim severity and improve the efficiency of claim-handling resources.