For example, they state that a Menninger Foundation study found that if a claim was not resolved within 60 days, the chance of the employee returning to work was greatly reduced. I suspect that there was more to this study than that one statement. In this instance, the time frame of 60 days may be an indicator of the severity of the injury, the overall fitness of the claimant, the quality of treatment being given or any number of other variables associated with the likelihood of successful return to work. Some injuries will take a year or more to resolve but the employee will be able to return to work. Other injuries will be so severe that the employee will not be able to return to work. Those injuries will take a great deal of time to resolve from a medical standpoint. You can't resolve a workers compensation claim until the employee has reached maximum benefit of treatment. So I fail to see how the 60-day mark is a predictor of anything other than a serious injury. Likewise, if the claimant has failed to return to work for 60 days and the claim handler is just realizing that there is a malingering problem, this is more of an indicator of a problem with the claim handler, who should have recognized the problem long before 60 days. Predictive modeling can do a lot, but it can't rectify poor claim handling.
Likewise, they state that by inputting the data about a claimant's use of medications and treatment history, the predictive modeling system will indicate that the claimant is over-medicating and over-treating. I seriously doubt that the system can say that. Predictive modeling indicates what is likely to occur, not what is certain to occur. So in this instance, the model could tell the adjuster that this claimant might be over-medicating and over-treating and that the claim deserves some scrutiny. My problem with this is that all claims deserve scrutiny. And at the outset of a claim, they all deserve the same amount of scrutiny. Once a certain amount of investigation has been done, it is possible to use the predictive modeling tool to help determine how to proceed further. However, a well-trained claims handler with a manageable caseload would also be able to determine that a claimant may be deviating from the norm when it comes to treatment.
It appears that the predictive modeling vendors are selling this tool as a means to shortcut the basic and thorough investigation that every claim deserves, in an effort to sell the product as a cost-cutting measure. I think this is a bad idea. It is far better to allow claims handlers to use predictive modeling as a tool to assist them in the course of the claim. It can help them identify fraud factors that they might not otherwise see. It can help them identify treatment that goes beyond the norm. It is still up to the claims handler, however, to recognize the underlying variables and investigate further to determine if there are legitimate reasons for the circumstance that the predictive model has flagged. Predictive modeling is not a substitute for solid technical claims handling skills and manageable caseloads.
While I do have some concern about how predictive modeling works, my bigger concern is with how insurers will implement it. In the not too recent past, some insurers were chastised by the courts for relying too heavily on reserving systems. These systems were meant to be a tool for claims handlers to use to provide them guidance in setting a reserve. Instead, the system became the basis for performance management standards. My fear is that predictive modeling will suffer the same fate.
Donna J. Popow, Esq., CPCU, AIC
Ethics Counsel and Senior Director of Knowledge Resources, American Institute for CPCU and Insurance Institute of