What’s in Claims Data?
The claims dataset we are exploring contains information on the injured employee such as
- Age, tenure, and wage information
- Type of injury, injury history, and medical prescriptions
- How long it took to report the injury and whether there is an attorney involved
When an employer makes a claim, there is an initial cost. For claims that have been closed we know the final claim cost. Taking the difference between the final cost and the initial cost gives us a notion of claim complexity. A large difference between the final and initial cost indicates a complex claim.
Since there are many pieces of information about each claim, it is difficult to investigate each one and figure out if there are certain patterns that make it more likely for a claim to be complex. This is where Tellius can help by leveraging machine learning techniques to tell us why some claims end up being complex.
Automated Analytics for Claims Data
Tellius offers a new breed of analytics that combines the power of machine learning with a user’s domain knowledge. To illustrate the power of Tellius, we can create a story (a type of automatic analysis) to better understand the characteristics of complex claims. The interface below shows how we can create a story to investigate what drives complex claims.
All that is required is to tell Tellius to investigate what causes complex claims. Taking this simple input to define a behavior present in the data, Tellius will automatically
- tell us which features in the data are the most useful to predict which claims will be complex
- show us different groups of claims that are often complex
- create a model to allow us to predict which future claims will be complex
Complex Claims Tell a Story
This story focused on complex claims shows several “segments” or groups of insurance claims that are very likely to be complex. For example, we can see that claims which
- take more than 5 days to be reported
- involve an opioid prescription
- and involve legal representation
turn out to be complex claims 89% of the time, which is 8.7 times more likely than the average claim.
This is one of many segments which are automatically learned from historical claims data. From here we can dive deeper to understand these segments. We can see which features have the largest impact on the likelihood of becoming a complex claim and investigate the interactions between these features. This is a much more efficient alternative to manual exploration of complex claims.
Understand and Operationalize ML Models
The output of this story allows us to understand different situations that lead to complex claims. We can also take action by making predictions about what will happen when a new claim comes in.
Behind the scenes Tellius has trained a machine learning model which can be used to predict how likely a new claim is to be complex. With one click, we can use that model to make predictions or set up a workflow to make predictions on a recurring basis. You can also create a customized model fine-tuned to your specifications. Taking this step to operationalize what you can learn from past claims data is the key to cost savings and more efficient claim processing.
In summary, Tellius provides a very user-friendly way to leverage the power of machine learning to investigate claim outcomes. With this increased exploratory and predictive power, insurance companies can become much more efficient in their claim processing.
Want to learn more?
Contact us here to schedule a demo to learn more about how you can use Tellius to improve insurance claim processing.