Built to catch what others miss.
NarcTrack's diversion detection engine is validated against a comprehensive suite of real-world clinical scenarios — before any event reaches your providers. We monitor every point in the chain of custody with the engine - giving you cradle-to-grave narcotic protection.
1,005
Test Cases
Across 63 clinical categories
100%
Pass Rate
Zero failures, zero regressions
26
Pattern Detectors
Independently validated
4
Risk Domains
Access · Inventory · Behavioral · Geographic
What we detect
26 validated pattern detectors
Every detector is independently tested against boundary conditions, real-world drug-specific parameters, and compound multi-pattern scenarios.
Access Domain
6 access pattern detectors
Off-Hours Access
Flags controlled substance movements outside a provider's established shift window, with graduated penalty by hours outside shift.
Access Domain
High-Frequency Access
Detects movement counts exceeding 2× a provider's rolling 30-day average, surfacing access patterns inconsistent with patient load.
Access Domain
Atypical Transfer Routes
Compares each drug movement origin-destination pair against the provider's established common routes. Unusual pairings are flagged.
Access Domain
Rapid Successive Access
Triggers when consecutive controlled substance accesses occur fewer than 20 minutes apart, with elevated severity under 10 minutes.
Access Domain
Shift-Boundary Activity
Monitors events within ±15 minutes of shift handoff times (07:00 and 19:00 ET), a known high-risk window for diversion incidents.
Access Domain
Chronic Shift-Boundary Pattern
Identifies providers whose events cluster near shift boundaries at a rate exceeding 30% of all events over 30 days.
Access Domain
Inventory Domain
5 inventory pattern detectors
Inventory Variance
Detects discrepancies between expected par levels and actual controlled substance counts. Schedule II drugs carry a 1.5× severity multiplier.
Inventory Domain
Waste Documentation Gap
Flags any controlled substance administration event missing corresponding waste documentation — a direct DEA compliance indicator.
Inventory Domain
Systematic Shrinkage
Identifies consistent, small-amount shortages across consecutive events — the hallmark of incremental diversion designed to avoid detection.
Inventory Domain
Micro-Shortage Tracking
Tracks sub-10mg individual event shortages (0.1–9.9mg) that fall below typical alert thresholds. Cumulative totals trigger elevated or critical flags at 50mg and 100mg.
Inventory Domain
Concentration Verification
Compares recorded drug concentrations against the organization's controlled substance catalog. Deviations greater than 5% indicate possible substitution.
Inventory Domain
Behavioral Domain
9 behavioral pattern detectors
Self-Witness Accountability
Detects when a provider records themselves as witness on their own controlled substance events — eliminating the independent oversight requirement.
Behavioral Domain
Chronic Self-Witness Pattern
Escalates when self-witnessed events represent 3 or more occurrences in a 30-day window, indicating a structural accountability failure.
Behavioral Domain
Pharmacodynamic Interval Validation
Validates that repeat administrations respect minimum safe intervals for each drug class — fentanyl (20 min), morphine (30 min), oxycodone (45 min), and more.
Behavioral Domain
Peer Statistical Outlier Analysis
Computes z-scores against org-wide 30-day drug usage baselines. Providers more than 2 standard deviations above peers are flagged; 3+ is critical.
Behavioral Domain
Sub-Therapeutic Dosing Detection
Identifies doses below 60% of weight-based minimums for fentanyl, morphine, midazolam, and ketamine — a key indicator of possible drug substitution or diversion.
Behavioral Domain
Prior Alert History
Incorporates a provider's 30-day alert history as a behavioral context signal, with graduated impact up to 40 points for 4+ prior alerts.
Behavioral Domain
Extended Custody Monitoring
Flags when a controlled substance box remains in a single provider's custody for more than 8 hours, with severity scaling beyond 16 hours.
Behavioral Domain
Chronic Sub-Therapeutic Pattern
Escalates when a provider exceeds 5 sub-therapeutic dosing events in 30 days, indicating a systematic pattern rather than an isolated incident.
Behavioral Domain
Peer Frequency Deviation
Benchmarks each provider's movement rate against the organization's mean and standard deviation, flagging significant outliers in access frequency.
Behavioral Domain
Geographic Domain
3 geographic pattern detectors
Location Anomaly Detection
Compares GPS coordinates of each drug movement against expected dispatch location. Events more than 500m off-location are escalated; beyond 3km is critical.
Geo Domain
Geospatial Impossibility Detection
Uses haversine calculations to identify physically impossible travel between consecutive GPS-logged events — a definitive indicator of falsified location data.
Geo Domain
Manual Override Monitoring
Flags events where location or system safeguards were manually overridden, providing a full audit trail of circumvention activity.
Geo Domain
How we validate
Rigorous by design
Every release of the diversion engine runs against the full 1,005-case validation suite. No deployment ships with a regression.
01
Boundary Condition Testing
Every numeric threshold in the engine is tested at threshold − 1, exactly at threshold, and threshold + 1. This covers 15 distinct boundary types across all detectors — eliminating off-by-one errors that produce false negatives at the exact edge.
02
Drug-Specific Parametric Sweeps
Each controlled substance has unique pharmacodynamic intervals, dosing guidelines, and scheduling classifications. Tests are generated per drug — fentanyl, morphine, hydromorphone, oxycodone, midazolam, ketamine, and more — with weight-adjusted dose calculations across four patient weight classes.
03
Compound Pattern Validation
Real diversion rarely involves a single flag. The suite includes 326 compound tests — 2-pattern through 6-pattern combinations — confirming that multi-signal events score and escalate correctly without pattern interference or double-counting.
04
Isolated Mock Data Injection
Detectors that depend on historical database state — prior administrations, peer drug usage, substance catalog concentrations — use injected mock fixtures. Each test runs against a known, controlled state, making results deterministic and reproducible across environments.
05
Clean Event Verification
Equally important: 28 clean-event test categories confirm the engine does not flag normal clinical activity. Tests span all controlled drug classes, mid-shift hours, standard routes, and correct dose ranges — validating specificity alongside sensitivity.
06
Risk Level Escalation Tests
The scoring model's four risk levels (low, medium, high, critical) and escalation rules are validated independently. Escalator conditions — such as 4+ high-severity patterns or patterns spanning 3+ risk domains — are tested to confirm correct floor behavior.
"The engine doesn't just detect diversion — it validates itself against 1,005 scenarios every time we ship. When a new pattern is added, the prior cases keep it honest."— NarcTrack Engineering
Test coverage
Every detector, by the numbers
Pattern detectors sorted by times exercised across the full validation suite. Each was verified to fire correctly and to stay silent when it should.
All 63 test categories — 100% pass rate
Every category from boundary conditions to real-world compound scenarios passed without exception on the current engine build.
Ready to see it in action?
The validation report is one part of the story. See how the diversion engine performs against your agency's real workflows.