Controlled Substance Suspicious Order Monitoring Modeling
A major pharmaceutical distributor retained Analysis Group to assist in building real-time order monitoring statistical models that identify suspicious orders of controlled substances and listed chemicals under the US Drug Enforcement Agency (DEA) Suspicious Order Monitoring Regulations (21 C.F.R. 1301.74(b)). The client requested that Analysis Group build a robust model that was tailored to the specific types of customers (e.g., retail pharmacies) they serve.
Analysis Group developed both a real-time model that could instantly determine whether a specific order was suspicious based on the customer's typical volume, frequency, and pattern, and also a set of retrospective automated analyses that could assist regulatory and compliance personnel in determining the appropriate action to take on the order or customer. Working with a multifunctional team consisting of legal, IT, regulatory/compliance, and business personnel, the Analysis Group team, led by Managing Principal Crystal Pike, and including Vice Presidents Pavel Darling and Kenneth Weinstein, analyzed more than two terabytes of historical order transactions to determine the most appropriate statistical techniques that aligned with past examples of suspicious orders.
The team made several recommendations:
- the client should adopt a dynamic, flexible model that relies on easily understood statistical techniques and provides clear output on why an order was deemed suspicious;
- the model should account for both ordering activity that is unusual relative to the specific customer's past ordering patterns, as well as ordering activity that is unusual relative to a peer group of similar customers; and
- the client should augment its Suspicious Order Monitoring by periodic analytical reviews of ordering and dispensing behaviors at the customer level to effectively monitor against diversion.
Our team worked on-site with the client to implement the model directly on their IT systems, where it now runs daily.