Infrastructure for Controlled Substance Diversion Detection

Controlled substance diversion detection requires documented definitions of what constitutes a suspicious pattern: acceptable waste rates per drug class, override frequency thresholds by role and unit, and peer comparison methodologies. DEA and Joint Commission standards mandate controlled substance policies exist and be current. These policies must be findable and enforced consistently—not interpreted individually by pharmacy supervisors—so the ML system applies uniform risk-scoring logic that can withstand regulatory scrutiny and HR/legal review of diversion investigations.

Controlled substance diversion detection requires automated capture of every ADC dispensing event, waste transaction (amount, witness, timestamp), medication administration record entry, and override event as they occur in real-time workflows. DEA regulations mandate controlled substance tracking already, making this comprehensive capture native to the environment. The ML system needs complete transaction logs with employee ID, patient assignment, shift schedule, and transaction timestamp to build behavioral baselines and detect anomalous patterns.

Diversion detection requires consistent schema across controlled substance transactions: employee ID, drug NDC, quantity dispensed, quantity wasted, patient assignment ID, shift schedule reference, and override indicator. All records must share these fields to enable peer comparison—comparing waste rates across nurses with similar patient acuity and drug assignments. The existing ADC and eMAR structured data provides this foundation, though clinical context (why a patient received higher doses) remains in narrative notes.

The diversion detection ML system must query ADC dispensing logs, eMAR administration records, employee scheduling systems (to verify patient assignment legitimacy), and HR records (role and unit assignment). API access across these systems enables the system to correlate a dispensing event against whether the employee was assigned to that patient on that shift—the core diversion detection logic. Controlled substance data access restrictions limit broader integration but don't prevent internal pharmacy-to-HR API connections.

Diversion detection baselines must update when staffing patterns change (new unit openings, role expansions), when formulary changes add new controlled substances, or when regulatory thresholds are revised. Event-triggered maintenance—new unit activation triggers baseline recalculation, new controlled substance formulary addition triggers monitoring rule creation—keeps the system calibrated to current operations. Peer comparison baselines require periodic recalculation as staff populations change.

Controlled substance diversion detection requires API-based connections between the ADC system, eMAR, employee scheduling/HR systems, and pharmacy management system. These connections enable the ML system to correlate dispensing patterns against patient assignment legitimacy, shift timing, and peer behavior without manual data assembly. The existing pharmacy-to-EHR integration provides foundation; HR and scheduling system APIs extend it to enable complete behavioral analysis.