Analytics: Predictions Vs. Presumptions

Plaintiff lawyers can teach us something about the limits of predictive modeling when it comes to workers’ comp claims processing. Simply stated, it is better to presume than to predict. Lawyer advertisements blatantly tap the mindset that employers, insurance companies and adjusters cannot be trusted to pay benefits. Notice that the advertisements do not have to prove this as factual because the advertisements correctly presume this notion is a societal norm. In fact, the mindset is so prevalent that the advertisements do not even have to depict a claimant’s current frustration, but only need to describe what might happen if claimants don’t hire a lawyer. Essentially, the ads address a void of employee confidence that causes concern even though there has not been any direct harm yet. Consider an injured employee who hires counsel right out of the gate. No advertisement is required. The employee is already in a “fight back” frame of mind. What makes that employee different from others who, weeks or months later, might be swayed by advertisement to retain counsel? The difference is that the employer had a chance to act on the employee’s natural concerns in a positive way but lost it. The lesson here is that lawyers do not need predictive analytics or predictive models to screen for complicated claims. Lawyers presume that every injured worker has doubts that can be transformed into feelings of pending injustice. Making the claim complicated is easy once they get the claimant roped in. Everybody knows that attorney representation often increases claim complexity and potential dollar value simply because the claimants often accept a dark notion of fairness and are willing to do more than what common sense and medical research supports to maximize their claim. I contend that there is a stark vulnerability in today’s industry reliance on analytics and predictive modeling. Automated models seek to assess new cases and save resources by assigning low-level indicators to a fast-tracked category. The models assume a claimant is emotionally fine. Instead, we should realize that every fast-tracked claimant is subject to lawyer advertising and cautionary comments by relatives or co-workers that agitate the claimant’s natural fears and suspicions. We must also accept the poor societal image of insurance adjusters as a reality. We all experience claims that start as medical-only, then turn bad. I contend that the aggregate cost of these missed opportunities obviates any argument that predictive modeling is good for WC claims. Quick Tip: Presume and Act, Don’t Predict and Wait Presume: Like plaintiff lawyers do, consider that each and every injured employee, even with the smallest injury, is a potential litigation candidate. They all have some degree of caution, low confidence, confusion and fear. Be First: Strive to be first in exposing and defusing even the most minor employee concerns. At the outset of an injury, all employees should know they will have an open forum and a direct line of communication in the course of their claim should any concerns arise. Take Responsibility: The employer, not the adjuster, should provide this open forum and line of communication. The employee must be confident because of some historical degree of trust established with the employer. The adjuster is simply not capable of creating that atmosphere. Straightforward Methods: Creating a proper forum is not complicated. Immediate meetings with relevant parties such as the supervisor, WC coordinator, human resources, safety, etc. should investigate the claim as well as alleviate any employee concerns. Nurse triage, as a vendor process, can add a powerful layer of assistance and confidence. Bottom line: It might take all of 15-45 minutes to make sure an injured employee feels his WC claim is being handled fairly and is very important to the company. This added effort is peanuts considering the cost of claims that go rogue. The Only Prediction That Counts: While claim vendors perpetuate the false notion of efficiencies in predictive modeling, you only need to make one prediction: that your injured employees need to be satisfied so they won’t be sucked in by presumptive lawyer advertisements.