Claim Analytics has been working on a number of projects, where the Claim Analytics predictive models are used to auto-adjudicate claims. Here are two examples:

Project 1

A Mid-West Workers Compensation insurer is using the claim scores to drive the selection for claims to be auto-adjudicated. Previously no claims were auto-adjudicated, and now they are running at 40%.

Claim Analytics built a predictive model using historic claims data to assess the risk of a claim incurring certain medical expense thresholds.

The claims are segmented into five categories:

• Auto 1 – fully auto-adjudicated, entitlement checked, payment approved cheque cut, regulatory reports completed. This represents 40% of their claim volume.
• Auto 2 – As above, with a manual review as part of payment approval process.
• Manual 1 – Claim routed via workflow to appropriate claim manager.
• Manual 2 – More complex claims with risk of medical expense explosion routed through more intensive review process.
• Manual 3 – For highest risk claims which are catastrophic or high likelihood of becoming very expensive.

The insurer took time to carefully audit the model’s scores and came to the conclusion that there was a 95%+ rate of agreement between what the model predicted and the outcome on the claim.

From the technology perspective, the insurer has an in-house claims management system, which accesses the Claim Analytics scoring algorithm as a module running alongside their own system, and is installed in their environment.

Project 2

For one of the largest LTD providers in North America Claim Analytics built a LTD claims approval model. This was built by looking back over their historic data going back 5 years and analyzing claims which were approved and claims that were denied.

Using this data, Claim Analytics built a model to predict the likelihood that a newly reported LTD claim would be approved. The model was built by quantifying patterns in historical approval decisions and applying this logic to score newly reported claims on the likelihood that they would be approved based on existing practices. The model was tested and shown to be an accurate predictor of approval.

Claims are scored from 1 to 3 as follows:

• A score of 1 will indicate that the cla
• A score of 2 will indicate that the claim need only check eligibility and compliance to the duration guidelines and then the claim can be approved
• A score of 3 will indicate that they must manually adjudicate the claim before approving it.

Conclusion

Claim Analytics believes both of these projects represent examples where significant advantage can be obtained with close integration of data analytics with claim management software. This has the potential to produce a highly sophisticated environment, with significant customer benefit.