Deterrence-Driven Algorithms for Audit Under the Sentinel Effect

Abstract

Fraud, waste and abuse are significant problems in major industries such as healthcare and manufacturing, particularly when third party payers such as Medicare are involved. Current practices for auditing fraudulent activity are based on scoring models used to select practitioners or claims that are likely to be fraudulent. These models ignore the “sentinel effect” that arises from the behavior modifications of other individuals who are aware of the audit. In addition to the direct benefits of auditing fraudulent individuals known as the audit effect, second order benefits are accrued due to this sentinel effect. Yet, current auditing algorithms do not take the sentinel effect into account. In this paper, we present deterrence-driven audit policies in the presence of audit and sentinel effects, and evaluate our algorithm using an analytical model and agent-based simulation. Our results indicate that significant reduction in healthcare costs can be achieved, while maintaining fairness, when auditing policies take sentinel effects into account.