carrot carrot carrot Change Centers x cognizanti collaborators create-folder Data Science Decisive Infrastructure download download edit Email exit Facebook files folders future-of-work global sourcing industry info infographic linkedin location Mass Empowerment Mobile First our-latest-thinking pdf question-mark icon_rss save-article search-article search-folders settings icon_share smart-search Smart Sourcing icon_star Twitter Value Webs Virtual Capital workplace Artboard 1

Please visit the COVID-19 response page for resources and advice on managing through the crisis today and beyond.


The (Data) Science Behind Preventing Insurance Fraud

The Challenge

Insurance fraud costs carriers an estimated $80 billion each year, with fraud in healthcare claims by far the largest culprit. A leading U.S. workers’ compensation insurance carrier suspected high levels of fraudulent claims. Like many carriers, the company used a rules-based claims review process that focused on individual medical invoices. Unfortunately, that process didn’t consider the context of past bills for either a specific claim or the provider filing the claims. Although historical data existed, it did not provide insight into fraudulent behavior patterns.

The company wanted to explore innovative approaches to this problem. It engaged Cognizant to develop a unique provider benchmarking methodology that would identify providers’ fraudulent behavior.

Our Approach

After developing and testing multiple hypotheses, Cognizant created an analytical methodology that uses unsupervised machine-learning techniques to gather and analyze medical bill data against three important dimensions: plausibility, outcomes and behavior. 

Our analysis evaluates the plausibility of providers’ treatment decisions using the entire gamut of data available in medical bills. To gain insight into outcomes, as well as monitor and control the cost and duration of patient treatments, we index data by categories such as providers’ diagnosis, state and specialty. When analyzing provider histories (behavior), our model flags anomalies including moves across state lines or frequent address changes, history of denied claims, atypical narcotic or opioid prescriptions and suspicious relationships with other providers that indicate collusion.

Machine Learning Injects Truth Into Medical Claims Analysis

Because fraud can occur across many areas of the claims process, machine learning is essential in mining the vast volume of data, detecting patterns in that data and accounting for nuances. Cognizant’s provider benchmarking methodology, which heavily leverages machine learning, directly prevents fraud to bring down all healthcare costs.


in fraudulent claims identified


return on insurance carrier’s investment


the efficiency and direction of special investigation units, claims research, clinicians and claims adjusters