Fraud Detection for Insurance and Medical Claims

EmcienPatterns reveals savings opportunities by identifying patterns of fraud not previously discovered with other techniques.

A large ensurer was facing millions in costs due to billing fraud. Their models’ ability to predict fraudulent billing activity had plateaued, but costs continued as fraudulent bills outpaced investigators ability to track and examine them. Analyzing the insurer’s historical claims data with Emcien, the insurer automatically detected all of the the complex combinations of rules that are predictive for fraud.

Fraud Predictions

Typically, rare outcomes like fraud are difficult to detect because they represent a very weak signal hidden in large amounts of noise. With Emcien, the insurer was able to identify every combination of events that identified fraudulent billing, increasing the accuracy of their fraud predictions by more than 10%. This transformed how the insurer targeted fraudulent claims investigations, representing a cost-savings of millions of dollars per quarter.

predict fraud in medical billing data at scale.

By automating the Emcien Prediction module, the insurer was able to implement a complete system of recurring analysis and predictions. Through this automated system the insurer’s predictions continued to adjust to changes in patterns of fraud. As tactics change, the automated system continues to track the patterns that are most predictive of fraud. Today the insurer continues to identify new fraud tactics without the hundreds of hours of manual analysis required by their previous methods.

Business Value:

  • 10% increase in ability to identify fraudulent claims.
  • An automated method that can keep up with changing fraudulent methods.
  • Integrates within existing systems via APIs to automate predictions into the workflow.

See a Demo

Want to see how this solution can help your team? Enter your contact information in the form below and an Emcien representative will reach out to show you a demo tailored to your use case.

See a Demo
Sending