This is the second in a set of three posts – you can see the first on Reducing the cost of manual claims handling here. The series is designed to use specific real-world stories to illustrate the importance of handling your claims effectively, not just efficiently. Digitizing your claims process may help you process a claim more cheaply, but you also need to ensure that the right claims are paid, and the right approach is taken for each. This means investing in digital decisioning – a proven way to reduce fraud and waste, assign the right people and improve loss ratios.
Business Problem
We’ve worked with many different kinds of claims across many different kinds of insurers. One particular project was around the effective handling of disability claims. Disability claims can be very costly to handle because they require a manually intensive process involving multiple visits and lots of documentation. Some claims can be “fast-tracked”, relying largely on emails and remote visits for instance, to reduce costs and improve customer service. For this insurer, the fast-track process was well defined and significantly cheaper. The challenge was defining which claims should be fast-tracked: Fast-track the wrong claims, and fraud or waste can slip through. Be too strict and potential savings would be lost. The business problem was not digitizing the fast-track claims process, but improving the expensive, inconsistent manual decision to identify suitable claims.
Solution
Fixing this problem involved automating this fast-track decision so it could be made rapidly, cheaply and consistently. And it meant doing this in a way that combined human and automated elements, leverage expensive external data appropriately and integrated exciting work being done by the machine learning team.
We began, as always, by developing a decision model for the fast-track decision. This model included both manual and automated elements, showing exactly how and when human judgment was appropriate. The sub-decisions that didn’t need human judgment were then automated using business rules, allowing a significant number of decisions to be made without human intervention and reducing the work effort when experts did have to be engaged.
This automation was used to assess the value of expensive external information sources. Evaluating the difference in automation with and without this information ended a long running argument, allowing the team that wanted to pay for the external information to prove its value and identify the right times to purchase it.
The decision model also allowed for the effective integration of machine learning for medical report analysis. A machine learning model was developed to identify how likely it was that a doctor’s report supported the claim by identifying the claimed condition as part of the medical assessment. This further reduced the need for manual review, allowing expert resources to be applied where the decision was more nuanced or complex.
Results
First and foremost, the team got an automated and consistent fast-track decision to use day to day. This maximized the rate of fast-tracking without exposing the team to more risk or fraud. It allowed external data to be bought when it would be useful and immediately applied to create value. Overall, the automated decision made a bigger difference to overall effectiveness than years of tinkering with the process had managed. Automating the fast-track decision was a game changer.
It also dramatically improved the time to value for machine learning. The decision model identified a clear role for machine learning model in the decision, focusing the data scientists on the specific prediction that would help. This more focused machine learning model was easier and quicker to develop than the one originally planned for. Furthermore, the decision model and the automated infrastructure provided guiderails and context that meant much lower accuracy was acceptable. The business owners could clearly see where the machine learning fit and were confident in their decision model, eliminating demands for excessive accuracy. A minimum viable model was developed, and operationalized, much faster than in previous attempts, getting to value quickly.
You can learn more about how digital decisioning can improve your claims handling by watching a recording of our CTO, Ryan Trollip, talking about claims automation with our partner Red Hat. As always, please do contact us with any questions about how we could help you with your process.