Discovering Provider Outliers in Billing Behavior

Detecting Outliers in Provider Claims to Reduce Insurance Fraud


Medical fraud and abuse accounts for an estimated $100-$170 billion annually*. Overbilling in the US healthcare system is unnecessarily increasing care costs and patient risks and decreasing payers’ ability to accurately classify the quality of providers.

Using advanced analytics techniques, EdjAnalytics works with insurers to discover provider outliers based on their billing behavior. By identifying providers with claims portfolios outside of the norm, Edj is able to illuminate which providers may be consistently overbilling or billing for unnecessary procedures. This knowledge will provide the basis to implement incentive programs, take actions to avoid unnecessary reimbursement costs from abnormal claims portfolios, improve medical care, and reduce patient costs.

Edj’s data science team leverages complex modeling techniques to discover anomalies in anonymized patient claims data. This approach allows the data to speak for itself, removing any assumptions or human biases. Successful techniques for pattern discovery include clustering methods and principal component analysis.

The results of the effort will be a tool to dynamically identify the pool of provider outliers.


  • Payer provides Edj access to their anonymized claims database.
  • Payer details the database contents, including the translation of procedural codes.
  • Edj develops engineered features from the raw claims data to facilitate modeling efforts.
    • Features may include age groupings, major/minor procedure groupings, diagnostic work, preventative work, anesthesia/sedation, procedure choice by cost, etc.
  • Edj performs analysis and modeling to identify outliers.
  • Edj provides a tool to showcase the outliers and highlights a few outliers as case studies to value the results of the work.


For every percentage point savings in reimbursement costs, providers have the potential to recover over $1.1M annually.

Estimated Project Timeline: Three months after complete data set is received

Contact Edj today to discover how we can help your business.

Source(s): * Rudman WJ, Eberhardt JS, Pierce W, Hart-Hester S. Healthcare Fraud and Abuse. Perspectives in Health Information Management / AHIMA, American Health Information Management Association. 2009;6(Fall):1g.