By: Jason Beans, CEO, Rising Medical Solutions
Predictive modeling is the latest hot term in the insurance industry, but what does it mean and what impact can it really have on outcomes. Let’s analyze what has been used historically and what predictive modeling promises to deliver.
First, know that the difference between predictive modeling and traditional reporting is that you are trying to find patterns that can help you take steps to improve outcomes early and proactively. In our industry this is crucial. When only 20% of claims make up 80% of the expense, we must find ways to intercept early and react swiftly to patterns and that’s precisely what predictive modeling does for us.
Here are some ways predictive modeling can be used in place of traditional methods to really make an impact on the bottom-line:
- Early intervention for case management
- Early pursuit of second medical opinions
- Comprehensive use of Utilization Review, gold standard treatment benchmarks, disability guidelines
- More accurate claim file reserving
- More positive involvement by the employer, (i.e. supportive return to modified work, more psychological “stroking” to make the injured employee feel valued to preempt much of the psychological and physical deterioration that occurs with prolonged disability)
- Consideration for additional investigative measures
- Consideration for early settlement before the claim “blows up”
- Team approach to claim handling for files that have issues identified during Predictive Modeling (i.e. multi-disciplinary roundtables, executive level, adjuster level, medical director, legal counsel)
All of these applications enable the user to react faster and move quicker, both keys to outperforming your competitors. Think of it as looking at historical data and finding patterns and indicators that produce bad (or good) results. Build predictive modeling to look for those indicators as early in the process as possible, even before a policy is written or an accident occurs. This can impact every area of insurance performance.
Just look at what banks use to evaluate a credit score. They weigh many factors that they have found indicate an ability and willingness to pay and use those predictive tools to produce a single score that tells them the safety level of a loan. The same holds true in what we do.
Recognizing and Reacting to Patterns
We had a client who had poor underwriting results in a state most carriers had great results. When we ran the historical data of agents and policies and overlaid them with crime statistics, the correlation was very strong. When we compared results in the low crime area to the high risk areas, the underwriting differences were dramatic.
Take those triggers and move them forward to even approving a policy or agent, and the outcomes are dramatically improved; add it to the policy or agency score. Factors like class code have always been used. Increase the indicators used in the calculation of the underwriting score to find the best risks to underwrite out of a class code. Other factors such as high risk zones for natural disaster and terrorism modeled in to help level and normalize concentrations of risk.
Evolving from Butcher to Surgeon
What I just described are simple broad predictive modeling techniques. Picture a butcher hacking at a piece of meat. Eventually, you want to evolve from the butcher to the surgeon. Refine and refine the indicators and metrics based on outcomes. Maybe only a few of those agents have issues. Perhaps many of the policies are worth writing in those areas, based on certain indicators. Maybe the risk only requires a slight increase in pricing and the predicted loss ratios will be adequate.
I was in a meeting with a carrier and their actuary voiced a concern over a recent dramatic increase in incurred claims payments. In looking at the numbers, Incurred But Not Reported (IBNR) had been adjusted down at the end of the year because the last three months of the year had less incurred payments than expected, and an extreme increase in claims payments over the first three months of the next year was indicating reserves were very light. Laying the data out by date of service verses data of payment, the data was much more level and predictable. It turns out there was a change in the PTO policy at the carrier and a high percentage of the mail room and claims staff were taking vacations at year end (bills were not getting paid) and they dug out at the beginning of the new year. The carrier put in place predictive reports based on date of service, staffing and other factors to prevent massive unnecessary adjustments in reserves and IBNR and improve their reserving results going forward.
Benefiting Claims Management & Medical Management
The area of claims management has infinite possibilities for better use of predictive modeling. Every experienced adjuster knows that there are alert factors that likely translate to a high cost claim. Time at the job, time between date of injury and reporting, the treating provider, high opioid usage, diabetes, obesity, etc. It’s not cost effective to put the same resources on every claim. Only a small percentage will drive the majority of the costs. Track early indicators that happen often and add them to a risk score so adjusters can identify and act on high risk claims. Do that well and achieve a huge difference in outcomes.
Everyone in compliance and underwriting knows the risk of stair stepped claims reserves. Underpriced clients, under filed experience modifiers, and regulatory issues all arise from ineffective reserving. With the right predictive modeling, it should be possible to track the impact of these indicators and weight them, achieving more accurate reserving with less stair stepping. Do not be afraid to start out as the “butcher.” After all, basic predictive modeling can still produce dramatically improved outcomes. Overtime, refinement allows you to evolve into a “surgeon.”
The medical management side will be changed dramatically by predictive modeling, and not just in workers’ compensation. Best provider by body part, nature of injury, best treatment patterns, etc. can be determined based on empirical, not anecdotal data. Move those best practices up front in the treatment plan and the outcomes will improve dramatically, reducing costs and improving employees lives.
Arming Yourself with the Right Tools for the Future
Proper predictive modeling requires some heavy lifting, particularly on the front-end. Historical data must be clean and complete. Workers’ Compensation is a long-tail item, and the reporting warehouses have not been around as long. We have seen many data sets missing claims expenses that result in inaccurate calculations. The first pass will likely require adjustments. In fact, continued refinement is the rule, not the exception. Even so, the power of predictive modeling for the insurance industry is infinite, and it should be part of any workers’ compensation payers’ future plans.
About Jason Beans
Jason Beans is the Chief Executive Officer of Rising Medical Solutions, a company he founded in 1999. Possessing more than 20 years of industry experience, Jason’s knowledge has been tapped by various media outlets including: The Wall Street Journal, CNN, NPR, Fox News, Chicago Tribune, LA Times, Business and Insurance, Insurance and Technology, AMBestWeek and Managed Care Magazine.
Prior to founding Rising, Jason held multiple executive positions in the medical cost containment field. He was the National Director of Property and Casualty Operations at Concentra, held several management positions with CRA Managed Care (eventually Concentra) and was a founding principal of BND Operations Company, a medical management consultancy. Notably, Jason is a longstanding contributor and consultant for the Rhode Island Workers’ Compensation Medical Fee Schedule rules, which Rising still updates and publishes today.
His expertise helps steer several Boards of Directors in the insurance and managed care industries and he is a frequent speaker on medical cost containment and MSA administration topics. Jason is a graduate of the MIT/EO Entrepreneurial Masters Program and he holds a Finance degree from Boston College.
About Rising Medical Solutions
Rising Medical Solutions (Rising) is a national medical-financial solutions organization that provides medical cost containment, care management and financial management services to the workers’ compensation, auto, liability and group health markets.
Based in Chicago, Illinois, Rising started as a two-person team and is now one of the fastest growing private enterprises in America, as ranked by Inc. magazine and the Private Company Index (PCI). To learn more, visit: www.risingms.com.