Lodestar Insights

As is the case with so many vended services in the workers’ comp claims industry, you’ll need to justify the expense against improved outcomes. Predictive modeling is no exception to this. Also, like most other services, it is hard to measure the negative outcomes that are avoided by taking guidance from a predictive model. However, there are certain actions prompted by predictive modeling that can be assumed to lead to more positive outcomes and reduced expenses.

Early Nurse Case Management Assignment

Assigning nurse case management early and on a set of claims identified by predictive modeling means better utilization of those services. It may also be that the number of claims predicted to need case management is less than the number currently being assigned.

Early Employer Intervention for Return to Work

It’s never too soon to talk about RTW with an employer. Especially when a predictive model indicates a potentially difficult return to work situation.

Early Medical Provider Interventions

Predictive models are very useful in identifying good and bad matches between medical providers and injured workers. Although there will always be constraints involved in the direction of medical care, adjusters and nurse case managers alike can greatly benefit from an early warning about a provider and injured worker mismatch.

Claim Denial

Predictive models can also identify claims that are good candidates for denial. A a successful denial is an extreme example of potential cost avoidance and a good demonstration of how predictive modeling is a key part of avoiding costs in workers’ compensation claims.

Another worthwhile observation is that all of the benefits that come from predictive modeling come early in the life of a claim. In other words, the earlier you act on predictive modeling guidance, the greater potential for more positive outcomes.

Although it seems like a simple question, this is perhaps the hardest thing to decide when exploring predictive modeling for your claims operation. The main thing to keep in mind is that you want to generate insights that cause someone to take action. Here are a few things that might be worth predicting and examples of the resulting actions they might prompt.

Basic Risk Level

Basic workers’ compensation claim risk could be described as a function of ultimate incurred. A nice way of assigning risk is to consider high risk claims as the top 10%-15% that generate between 75% to 90% of of your total incurred costs. This kind of prediction requires that it be performed early in the life of a claim so that adjusters and managers alike can focus on all the possible actions that could prevent predicted high risk claims from becoming actual high risk claims.

Claim Reserves

A cousin of predicting claim risk would be trying to identify how reserves should be set or adjusted for a given claim. Actions taken from this prediction are fairly self evident in that reserves can be adjusted according to a blend of adjuster experience and model guidance.

Potential For Return to Work

Failed Return to Work (RTW) is a killer for workers’ comp. claims. Being able to identify a potentially challenging return to work case prompts early intervention with injured workers and employers alike. Alternatively, this data in the hands of good nurse case management can help mitigate the instance of lost work days which often lead to complex return to work cases.

Of course, there are many other things worth predicting with respect to workers’ comp. claims. The keys are to be clear in what you want to predict and to try and predict it as early as you can.

There is a lot of buzz around the use of Big Data in Insurance, and particularly, the workers’ compensation claims industry. Although there is no doubt that this is a complicated field, let’s take a moment to cover some of the basics of what predictive modeling is, what it isn’t, and what it means for workers’ comp claims.

What Predictive Modeling Is

Whether you call it machine learning , analytics, or data modeling, you’re referring to the process of using historical data to make future predictions. A significant part of building a predictive model includes data mining. The art and science of data mining, practiced by data scientists, involves finding patterns and stories in data that explain real world phenomenon. Sometimes the data used to support a story or pattern is obtained from an outside source and it is the relationship between that data and your own that really powers your predictive model.

What Predictive Modeling Isn’t

Predictive modeling isn’t a crystal ball: past performance doesn’t always guarantee future outcomes. Although, if you’ve got a good enough model using the right data, it usually does. Predictive modeling (especially when referred to as artificial intelligence) is also not an adequate replacement for human intuition and know-how.

Applying Predictive Modeling to Workers’ Comp Claims

By implementing predictive modeling in your operation through either your claims or risk management system you can begin to effectively:

• Inform adjusters about possible claim issues or outcomes
• Allow for management to better triage claims and manage workloads
• Focus C-Suite and client attention on the claims that currently do, and likely will, matter

Assuming these are the goals you’re looking to achieve, your next step is to look at your data and find a platform that can execute your model.