By Cory Bilton
Numerous factors go into determining the value of an injured client’s case. Initial injuries, cost and duration of medical care, long-term injuries, and the limitation of a person’s activities are just a few of the factors to consider in determining the value of a personal injury claim. How about the injured person’s FICO credit score? It’s not a factor that we consider. However, the insurance industry is adopting new methods to search for just this kind of unconventional connection. They call it predictive modeling and it is Big Data’s latest promise to the insurance industry. Since most personal injury cases involve insurance adjusters and defense attorneys hired by insurers, predictive modeling will have a significant impact on the future of personal injury law.
What is Predictive Modeling?
Predictive modeling is the process of using statistical methods and large sets of data to look for hidden connections in the data. The idea is that the hidden connections may be useful predictors of an individual customer or claim outcome. For example, an insurer can collect data on an individual driver through a telematics device, look for predictors of the likelihood of claims within that data, and then set the price of the policy. The same method can be applied to claims too. As data becomes easier to collect and analyze, insurers are learning to search for anything that would allow them to predict the outcome of claims quickly or reduce costs.
Haven’t Insurers Always Looked for Predictors?
Since the very first insurance agreement, insurers have tried to predict the future to price the policy. Usually the information the insurer uses is easy to collect or is intuitively obvious. For example, auto insurers have long considered the amount of property damage to the car as a predictor of the severity of the driver’s injuries. Predictive modeling aims to dig deeper. Instead of just focusing on obvious predictors, predictive modeling calls first for collecting a huge pool of data about your users. Newly developed statistical methods are then applied to find the connections in the data. The predictive factors are not initially known. Instead, they are discovered from the huge pool of data itself. The goal is to discover the less obvious, or hidden connection between a piece of information and an insurance claim outcome.
How Widespread is Predictive Modeling in the Insurance Claims Process?
Predictive modeling has developed alongside other Big Data analytics. For the insurance industry, it’s a new tool that is growing in popularity. A recent survey by Towers Watson found that 48% of large insurers “are either using, are in the process of using or are nearly ready to roll out predictive analytics for fraud detection . . . .” Further, 26% of large insurers are already using, or close to implementing, predictive modeling to triage claims for adjuster assignments. By comparison, 80% of large insurers are using predictive modeling in pricing, risk selection, and underwriting. So it’s likely that predictive modeling will become a bigger part of the claims process over time.
How Will Predictive Modeling Effect Personal Injury Law?
1. Settlement Offers Will Become Less Predictable
Predictive modeling is all about the hidden, less obvious predictors of claim outcomes. Thus, an insurer’s decisions will become increasingly opaque as it implements predictive models. In fact, newly found predictors in the claims process are likely to be closely held trade secrets by insurers. This will make personal injury cases less predictable. This unpredictability might not always be on the low end; some initial offers may be higher than expected. For example, if a personal injury case has a hidden factor that the insurer has identified as being a high-cost claim, then the insurer may offer more up front in an attempt to settle the claim quickly. But on the other hand, if an injured person’s claim happens to possess the hidden factors that the insurer has determined indicate fraud, the adjuster may respond with extra suspicion and a low offer. Until these hidden predictors become known among personal injury lawyers, each side will fail to see eye to eye more often.
2. More Cases Will Result in Litigation
Juries won’t give much weight to esoteric statistical connections discovered by predictive modeling. Even if the hidden connection is statistically demonstrable, if it defies common sense defense attorneys will not make the argument at trial. Therefore, while pre-litigation claims maybe increasingly controlled by statistical models, once a case goes into litigation, predictive modeling will go out the window. Jurors are human. If the hidden connection resulting from a predictive model is too hidden for intuition, then it’s not going to help sway a jury. Therefore, initially at least, the antidote to predictive modeling is litigation.
3. New Public Policy Concerns Over Privacy and Information Use
What happens if predictive modeling finds a connection between the claimant’s religious beliefs and claim outcome? How about a connection between a claimant’s race or national origin and claim outcome? What kind of information may an insurer consider and what kind of information is off-limits? With statistical models that specifically attribute certain information as predictors of claim outcomes, should we, as a society, forbid the consideration of certain predictors? Are there certain types of information that are constitutionally protected? These are weighty questions. Since most insurance regulation is governed at the state level, these decisions will initially occur there. One state may permit the use of predictors that another state prohibits. As data becomes easier to get, can personal injury lawyers prevent certain information about their clients from being divulged? In this area, I can only posit questions, but it seems quite certain to me that these will be issues for public discussion.
Predictive modeling is happening. Insurers are finding new ways to use it each day. We may discover new determining factors to value personal injury cases. Or maybe not. Maybe we find that predictive modeling just produces a lot of correlations and no causation. But predictive modeling’s long term use, and usefulness, is going to require more discussion and debate. Using hidden factors to determine the value of an injured person’s case may ultimately cause more harm than good. But predictive modeling’s current promise to insurers is too good to pass up.
Please check out my disclaimer.