These are prevention and detection programs, not enforcement cases. A program is listed only if its result was published with numbers and the source was fetched and read on 1 November 2026. Two sources are an abstract or a session listing rather than a full paper; their profiles say so. Where a source leaves out a detail, the profile says that too. For people caught diverting, see the diversion case studies.
Peer-reviewed journal
Fraser Health Authority, British Columbia
A dedicated lead plus routine software review found far more diversion than the old approach did.
- Organization
- Regional public health authority (Canada)
- Size
- 12 acute care hospitals and other facilities; serves a region of 1.9 million people
- Year
- Started 2019; results through December 2023; published 2024
- What they did
- Hired a pharmacist as Drug Diversion Lead to run the program.
- Formed an oversight committee (pharmacy, nursing practice, human resources, risk management, security) that wrote a diversion policy and an investigation protocol.
- Reviewed automated dispensing cabinet (ADC) diversion-software reports three times a year for every unit with an ADC, then audited charts for flagged staff.
- Ran each investigation with a standing team and a written final report to leadership.
- Measured result
- 1 confirmed diversion in 2018, the year before the program. 13 investigations and 8 confirmed in 2019, the first year.
- January 2019 to December 2023: 45 investigations, 23 confirmed (51%).
- 24 investigations (53%) started from staff reports and 21 (47%) from software.
- 40 of 45 investigations (89%) found practice deficiencies, whether or not diversion was confirmed.
- Limits
- More confirmed cases is the expected sign of better detection, not of more diversion. Single author, descriptive case study.
Source: Braun N. Establishing a Multisectoral Collaborative Drug Diversion Program in a Canadian Health System. Can J Hosp Pharm. 2024;77(3):e3551.
pmc.ncbi.nlm.nih.gov
· checked 1 November 2026, HTTP 200
Health system annual report
Kaiser Permanente San Jose Medical Center
Daily visibility and a 24-hour rule cleared a standing backlog of open variances.
- Organization
- Medical center in an integrated health system (Northern California)
- Size
- Not stated in the source
- Year
- Process started October 2022; huddle reporting from July 2025; result reported in 2025
- What they did
- Ran a process called Control Check that reconciles dispenses, administrations, wastes and returns using EHR (KP HealthConnect) and Pyxis data.
- Required diversion education for all staff with controlled substance access.
- From July 2025, reported weekly variance data at the daily safety huddle.
- Set a rule that every open variance is resolved within 24 hours. Nurse managers worked each one with pharmacy and frontline staff over the next 2 to 3 months.
- Measured result
- Open variances went from about 30 to 40 a day to zero.
- The source says zero has held for months. It does not give a number of months.
- Limits
- Self-reported in the organization's own nursing annual report. No unit-level data or bed count published.
Source: Kaiser Permanente Northern California. A model for safe and transparent controlled substance management. 2025 Annual Nursing Report.
nursingncal.kaiserpermanente.org
· checked 1 November 2026, HTTP 200
Conference session abstract
Methodist Le Bonheur Hospitals North and Germantown
A process improvement project cut unresolved discrepancies by about a third.
- Organization
- Two hospitals, pharmacy department
- Size
- Two hospitals; bed count not stated in the source
- Year
- Presented May 16, 2024; project dates not stated
- What they did
- Ran a multi-step process improvement project aimed at two things: fewer unresolved discrepancies, and more controlled substances dispensed from ADCs.
- The published listing does not detail the individual steps.
- Measured result
- Total unresolved discrepancies fell 36% over a six-month period compared with the previous year.
- Narcotics dispensed from ADCs rose 64% in one year.
- Limits
- The only public source is the session abstract for an IHFDA webinar given by the pharmacy director. The underlying data are not published.
Source: Gandy A. Addressing Unresolved Controlled Substance Discrepancies and Increasing the Number of Controlled Substance Dispenses from Automated Dispensing Cabinets. IHFDA webinar, May 16, 2024.
ihfda.org
· checked 1 November 2026, HTTP 200
Peer-reviewed journal
Stony Brook University Hospital, New York
Putting a cabinet and waste bin where clinicians finish cases shortened the time they carry used opioids.
- Organization
- Academic hospital, anesthesia department
- Size
- 788 beds; 25 main operating rooms; about 36,000 procedures a year
- Year
- Published August 2024
- What they did
- Added a Pyxis cabinet with an adjacent controlled substance waste bin in the post-anesthesia care unit, closer to the end of the case than the original machine in the main OR.
- Measured result
- Self-reported time carrying used narcotics dropped from a median of 10 to 20 minutes to under 10 minutes (p < 0.0001), among 78 anesthesia clinicians analyzed.
- Waiting until the end of the shift to waste also dropped (p = 0.0002).
- Limits
- Based on a recall survey, not audited waste records. Single center. Off-site locations were excluded.
Source: Parikh S, et al. Pyxis Proximity Improves Narcotic Waste Compliance Among Anesthesia Personnel. Cureus. 2024;16(8):e66800.
pmc.ncbi.nlm.nih.gov
· checked 1 November 2026, HTTP 200
Peer-reviewed journal
Academic hospital anesthesia department (Philadelphia)
Showing providers a live reconciliation report cut the error rate from 8.8% to 5.2%.
- Organization
- Academic hospital, anesthesia department
- Size
- Not stated in the abstract. Authors were affiliated with Thomas Jefferson University and Thomas Jefferson University Hospital, among others
- Year
- Data January 2012 to September 2015; published June 2016
- What they did
- Captured every ADC controlled substance transaction in near real time and compared it with the anesthesia record.
- Gave anesthesia providers a reconciliation report, refreshed every minute, at their workstations: patient, provider, ADC balance, charted total, and the difference.
- Measured result
- Reconciliation error rate: 15.8% at baseline, 8.8% with next-day reports, 5.2% with near-real-time reports (difference -3.6 points; 95% CI -4.3 to -2.8).
- 50% of discrepancies came from 12.4% of providers.
- Limits
- Taken from the PubMed abstract; the full paper is paywalled. Before-and-after design at one institution. Most errors were clerical, so this measures accountability, not confirmed diversion.
Source: Epstein RH, et al. Controlled Substance Reconciliation Accuracy Improvement Using Near Real-Time Drug Transaction Capture from Automated Dispensing Cabinets. Anesth Analg. 2016;122(6):1841-55.
pubmed.ncbi.nlm.nih.gov
· checked 1 November 2026, HTTP 203
Peer-reviewed journal
10 hospitals in 4 health systems, including Piedmont Healthcare and Scripps Health
Combined data and a trained model flagged known diverters months before existing methods did.
- Organization
- Multi-system retrospective study
- Size
- 10 acute care hospitals; 27.9 million transactions by 19,037 nursing, 1,047 pharmacy and 712 anesthesia clinicians
- Year
- Published 2022
- What they did
- Combined data from several hospital IT systems into one dataset.
- Trained a supervised machine learning model to classify each medication movement as low or high risk of diversion.
- Ran the model on 8 to 24 months of history containing 22 known diversion cases, blinded, and compared detection dates with when each case was actually caught.
- Measured result
- All 22 known cases were detected earlier than existing methods: mean 160 days, median 74 days earlier (range 7 to 579 days).
- On the initial sample: 96.3% accuracy, 95.9% specificity, 96.6% sensitivity.
- Limits
- A retrospective test on historical data, not a live program. The first two authors work for Invistics, the software company whose analytics were tested.
Source: Knight T, et al. Detecting drug diversion in health-system data using machine learning and advanced analytics. Am J Health Syst Pharm. 2022;79(16):1345-54.
pmc.ncbi.nlm.nih.gov
· checked 1 November 2026, HTTP 200
Numbers are quoted as each source reports them and are not comparable across programs: baselines, time frames and measures differ. Know of a program with a published result? Send it through the contact form with the source link.