Making Predictive Analytics Actionable
Imagine how powerful a tool would be that could see into the future. While such a tool may not yet exist, insurance companies have been using their unprecedented access to data to build predictive models for a number of years now. The use of big data can help identify claims that are likely to develop in an adverse way. This is quickly becoming a best practice within our industry.
With access to such wonderful and valuable information, there are still potential issues. The issues lie in how organizations operationalize – or more specifically, fail to operationalize – predictive analytics.
A lot has been written about the importance of data integrity, data mining and the identification of the most predictive attributes to factor into these algorithms. However, what tends to get overlooked is how organizations will use the output of these models within the claims organization to make better and faster decisions. The most common reason these kinds of initiatives fail is not necessarily bad data, but bad implementation.
If a claims department doesn’t use the output from their predictive modeling to change how they are managing their claims, it won’t drive a significant improvement in their outcomes and is likely to fail. It’s not enough to have a successful model that will provide for the early identification of claims likely to develop into large losses. Claims leaders need to think through how they are going to use that information to change the trajectory of potentially severe claims. That is the key to success.
When implementing a claims predictive model focused on early identification of large losses, a well thought out plan as to how this will be operationalized should include at least the following elements:
- New claim intake
- Staffing model
- Claim assignment
- Return to work
- Auto adjudication
- Settlement strategy
- Litigation management
- Key metrics/outcomes reporting
- Fraud identification
A good approach is to look at each of these key areas and ask the question: “How would we design this if we were starting from scratch with the advantage of predictive modeling, versus trying to simply tweak what is already in place?”
If a claims department doesn’t use the output from their predictive modeling to change how they are managing their claims, it won’t drive a significant improvement in their outcomes and is likely to fail.
Let’s take new claim intake for example. This is an area where insurers can use their analytics efforts to identify certain new data elements that could be collected early in the claim process, yet are shown to be highly predictive of claim development. By collecting the right data at intake, potential large claims can be identified sooner which will in turn enable more impactful triage and facilitate claim assignment to the right resources earlier in the claim life cycle. By getting the right claim and medical resources involved sooner, we can see improved recovery and return-to-work times along with reduced average severity.
Investing the time and resources necessary to operationalize the model up front will result in a much more successful implementation that will drive improved outcomes for your company.