Making workers’ compensation medical costs more manageable

  • Print
  • Connect
  • Email
  • Facebook
  • Twitter
  • LinkedIn
  • Google+
By Rong Yi, Stephen R. DiCenso | 21 June 2012

Workers’ compensation insurers can now tap newer, more advanced predictive modeling platforms as a way to contain the spiraling costs of medical claims and thereby gain a competitive advantage. But they can only take advantage of these modeling platforms if they have endorsement from all levels of the organization, from claims adjusters up to the chief executive officer.

This “buy-in” might be easier to come by these days, as the landscape has continued to change rapidly. The medical portion of workers’ compensation plans is increasing much faster than other workers’ compensation costs or general health care costs. Workers’ compensation medical costs have increased by 6% annually from 2002 to 2011, according to the National Council on Compensation Insurance. Yet both workers’ compensation indemnity payments and the medical consumer price index have each risen by just over 4% over the same period. Workers’ compensation medical costs have grown to 59% of total losses in 2011, up from 52% in 1999.

Some larger insurers and third-party administrators (TPAs) have recently begun to tackle rising medical costs by implementing claims predictive models. Companies have found that these next-generation models enable changes in claims strategies and culture. The key strategic changes are focused on:


  • Identifying and estimating future high-cost claims earlier.
  • Identifying and quantifying the cost (risk) drivers of these claims.
  • Creating a focused, early-intervention medical management program to prevent adverse claim development.

The good news for mid-market insurers and smaller TPAs is that more powerful predictive modeling tools have come to market, and at good price points. With increased computing power, these tools are now more accurate in helping identify problem claims.

To be sure, claims adjusters are doing what they can with the tools at their disposal. Predicting individual claim outcomes is what claims adjusters do every day. They navigate the complexities associated with each injured worker’s unique situation. Industry caseloads for workers’ compensation claim adjusters average 150 or more lost-time claims, so attending sufficiently to every injured worker’s claim is certainly a very daunting task. Even the best adjusters cannot prevent some claims from deteriorating beyond what a best practice outcome would suggest.

In addition, insurers today are confronted with shifting demographics and disease trends, such as an aging workforce, and obesity-related illness. When something operational or environmental is new and different, insurers need to quickly determine how the changes in the composition of their claims portfolio will affect claim costs.

There’s the old adage in insurance that 20% of claims are responsible for 80% of losses. We believe that the best way to assist adjusters in the claims administration process is to help them process more information and create a dedicated focus on the 20% of claims that have high medical expenses.

Predictive modeling can help workers’ compensation insurers find red flags, guide their claimants toward more effective and less costly treatments, and even detect potentially fraudulent claims. Insurers that have made predictive modeling most effective have embraced change in four key areas: data collection, the model itself, operations and culture, and business value. The following case study illustrates how one company was able to make changes in these areas and gain value from enhanced predictive models.

A large retailer revamps workers’ compensation

A large retailer that self-administers its workers’ compensation program has more than 1,500 lost-time claims annually. To get the most out of the modeling, the retailer embraced the four key areas of change.


Data as a foundation

The first step was to see what the client had in terms of data. Companies are in various stages of creating databases that can provide for predictive-modeling analytics. We worked directly with the medical bill re-pricing vendor to merge claim data, medical transaction data, prior claim data, and other metrics. External data was excluded to keep the focus on the client’s own medical-management outcomes.


Predictive model

Many high-cost claims did not appear to be very different from other claims in their early stages. However, as time went on, certain characteristics and patterns started to develop that increasingly differentiated them from other lower-cost claims.

To address this cost pattern, a model with an ability to monitor costs over time was needed. Predictive analytics were developed for the model at various points in time in the life of a claim.

To capture the multitude and complexity of claim characteristics, as well as medical cost and utilization patterns, we explored a large number of multi-dimensional interactions among these factors. We tested their statistical significance and predictive value at various points in time. For example, medical utilization patterns, such as seeing more providers, the use of advanced imaging tests, and pharmacy usage—including specialty drugs administered in a physician’s office and drugs treating stress, anxiety, and depression—were increasingly influential over time in determining future medical costs.

The model showed that different factors, or “risk drivers,” are at play, depending on how advanced the claim is. Predictions would change based on how much time had elapsed. And the models’ predictive accuracy increased as claims matured, clearly showing which cases would likely go bad. The risk drivers were integrated into an interactive reporting tool that scored and ranked the open claims in terms of predicted future payout. The tool compared the model predictions to actual case reserves, and identified the risk drivers for why each claim was expected to have high future payouts.

The risk drivers break down the predicted medical costs into the following categories:


  • Claim characteristics
  • Medical conditions
  • Medical utilization
  • Prescription drug categories

Many of these factors validated claims management intuition, while others demonstrated trends that would not otherwise have been known to most adjusters. All in all, it was viewed as a crucial supplement to the existing claims administration process.


Operational and cultural shifts

In order for the predictive model to work its magic, it needs to become an integral part of the claims operation. There must be an operational and cultural shift in the company. This is the only way to derive the model’s full benefit.

Our client had previously established a solid foundation for handling claims, such as a vigorous return-to-work program, triage protocol at key facilities, and strong safety and loss-control programs. With limited claims adjuster resources and lots of claims and medical transactional data at hand, the client was from the start looking for a breakthrough solution to further reduce workers’ compensation medical costs.

To engage the client in the predictive modeling process, a workshop was held with a multi-disciplined team of experts, including claims management, claims adjusters, a medical director, a nurse consultant, attorneys, and performance analytics professionals. The workshop resulted in the identification of numerous cost drivers and facilitated the recommendation of specific business actions and rules to reduce costs. It also led to a reworking of claims staffing structures, improved coordination with medical management providers, and the introduction of new data fields that would be collected within the claims administration process.

Just as important was how the workshop generated widespread enthusiasm, the development of a consensus view for the need to change, and a curiosity regarding potential enhancements of the claims-administration process.

Impressed by the value of the merged data, the client concluded that gathering additional information when a claim is first reported would better help identify potentially costly claims. Given that this information was not currently in a structured format, the client decided to invest in capturing more information and improving the quality of existing data by implementing a revised claimant questionnaire at first report of injury. This questionnaire included about a dozen questions inquiring about the physical and mental health of the injured workers. This data will ultimately be codified and made available as input into the predictive model.


Building long-term value

An excellent estimated return on investment provided final validation of the company’s decision to ensure it has the best possible focus on controlling the future costs of its workers’ compensation portfolio. The cost of implementing predictive modeling projects pales in comparison with the expected savings in medical costs.

Scoring and ranking claims—and providing results to the claims department—is not the end game, it is just the beginning. The efforts involved are indeed transformational, using powerful analytics to drive operational and cultural change in the claims organization. Predictive modeling will bring value to any claims department willing to embrace a new process—and the cultural and operational changes that are also needed to make it all happen.