Applying Aster’s Capabilities > Healthcare > Detecting Fraud

Teradata Guided Analytics

Aster is an analytic platform upon which we have built many Applications in almost every industry. The field of Healthcare is particularly rich with opportunities up and down the delivery chain. In this post I will focus on one of the downstream applications that we developed to leverage behavioral analytics to identify potentially fraudulent physician activity.


Abuses of the Medicare/Medicaid reimbursement program are splashed across the news headlines with regularity as prosecutors bust individuals or rings of unscrupulous caregivers. The abuse can take many forms ranging from straight up false claims submissions (billing for procedures not actually performed) to much more elaborate “creative” classification of care performed. Other forms of abuse include excessive referrals for specialized treatments to other physicians who make reciprocal referrals to churn maximum reimbursements from any patient who enters their medical offices.


While it is difficult to arrive at an exact amount of fraud that occurs, estimates are around $65 billion per year. The problem is so large that the US Office of the Inspector General has established a strike force whose sole purpose is to ferret out and publicize the fraud they uncover. A constantly updated “scoreboard” of the strike force’s enforcement actions (including a running tally of dollar amounts) is posted online for all to see.


The application developed by the Aster Solutions Team detects suspicious actors in the system by collecting behavioral data (e.g. prescribing activity, referral activity, treatment activity, etc.) and establishing a baseline expectation for these. Once these are established, approaches such as vector similarity or pSalsa are used to identify outliers who are likely padding their billing in some way. Additionally, our application leverages SQL-GR to reveal the network of interactions in the data to illuminate clusters of affinity between practices that are likely to be rings of coordinated fraud.


The data available when we built this Application was limited to the publicly available CMS Part B database of reimbursement requests, but a richer data set of activity would provide a more complete picture and yield even better results.


If you are interested in exploring any of the potential applications of the capabilities in Aster in the area of Healthcare, please contact me to engage with the Aster Solutions Team.