Sentri7 Drug Diversion

The determination is the product. Sentri7 Drug Diversion exists to weigh more than 60 risk determinants across more than seven transaction sources and return a judgement about which behaviour warrants investigation, and the vendor names machine learning as the method in its own voice rather than hiding behind the word analytics. Remove the model and nothing sellable remains, which is the test this axis applies and which the dispensing estates in this lane do not meet. The record is scoped to the diversion module specifically, so the grade is not diluted by sibling products that would not qualify on their own.

The intended division of labour is stated more explicitly than most, and it is anchored to something outside the company. The published study frames the goal as automating detection work in order to return pharmacist time to patient care, and cites the American Society of Health-System Pharmacists position on artificial intelligence in pharmacy as the standard it is working to. Citing a professional body's stance rather than asserting a private one is the right instinct.

What is missing is the operational layer: no published escalation threshold, no stated qualification for the person reviewing a flagged clinician, and no description of what the reviewer is shown about why the score arose. Recent trade reporting on this category, which names this product and its closest competitor, identifies staff training and trust in the alerts as the practical constraint rather than detection capability, and that reporting sits behind a subscription so it could not be read in full here and is recorded as a lead to follow rather than a finding.

The most transparent record in this lane. The method is named as machine learning and advanced analytics, the input surface is enumerated as more than seven distinct data sources with each named, the model is described as weighing and combining more than 60 determinants of risk, and the performance is reported in a peer reviewed venue with accuracy, specificity and sensitivity given separately rather than collapsed into a single marketing figure.

It falls short of the top grade because the determinants themselves are counted rather than listed, the model family and training procedure are described only at the level a journal abstract permits, and the published characterisation dates from 2022 with no subsequent update located for the current build.

No model provider, framework, library or third party analytics component is named for the machine learning underlying the risk scoring, and no distinction is drawn between capability built by Invistics before the acquisition and capability added or rebuilt by the parent since. That second gap is specific to this record. A buyer reading the 2022 evidence has no published means of establishing how much of the system that produced those results is still the system being sold.

The only peer reviewed efficacy evidence in this category. A study in the American Journal of Health-System Pharmacy, funded by the National Institutes of Health, evaluated the software across ten acute care inpatient hospitals in four health systems, more than 20,000 clinicians and over 25 million medication movement transactions, and reported detection of all 22 known diversion incidents at 96.3 percent accuracy, 95.9 percent specificity and 96.6 percent sensitivity, with known cases surfaced a mean of 160 days and a median of 74 days earlier than existing methods.

That is a defined population, a stated gold standard and a measured result, which no competitor here offers. Two qualifications travel with the grade. The study assesses the vendor's own software, so independence should be checked against the authorship and funding declarations rather than assumed. And it examined the platform as Invistics operated it before the 2023 acquisition, so it describes the product's lineage rather than certifying its current build.

Whether customer transaction data trains models that persist beyond the tenant, whether risk weightings are fitted per customer or across the installed base, and what retention applies to behavioural records about named clinicians are all unaddressed in published material. The question has unusual force for this product because its strength is breadth of input.

Employee time clock data, wholesaler records and retail pharmacy transactions are being joined to build a behavioural picture of individual staff, and the more sources a model reconciles the more consequential a silence about data reuse becomes.

Two passes surfaced no business associate agreement position, no data ownership statement and no description of how protected health information is handled by a product whose entire function is reading identified clinician and patient medication records. The parent demonstrably knows how to publish this material, since it maintains a detailed product level security and data protection page for its tax and accounting software. Nothing equivalent exists for the healthcare surveillance line, which makes the absence a choice rather than an oversight.

Two passes found no certification, attestation, trust portal or security page for this product or for the healthcare surveillance line it belongs to. The finding that makes this a clear grade rather than a marginal one is internal to the company.

Wolters Kluwer publishes a detailed product level security page for its tax and accounting platform, stating an ISO 27001 certification, naming the hosting provider, describing alignment to the National Institute of Standards and Technology cybersecurity framework and setting out availability and disaster recovery commitments. The product that reads controlled substance records and identified clinician behaviour has no equivalent. The capability and the template both exist inside the business and have not been applied here.

This is surveillance software rather than a regulated device, so no clearance applies, and the product is built around helping customers meet obligations to the Drug Enforcement Administration and state boards of pharmacy, with audit and reporting output positioned as defensible evidence. Regulatory context is therefore engaged rather than ignored, which separates it from vendors that leave the subject alone entirely.

It stops short of a higher grade because the company publishes no statement of its own regulatory position, meaning what claims it does and does not make, and no guidance on how its output should and should not be used in a disciplinary or law enforcement process that a diversion finding frequently becomes.

This vendor does one thing on this axis that nobody else in the lane does: it published the specificity figure. At 95.9 percent specificity across 25 million transactions, a buyer can work out roughly how many transactions get wrongly flagged, and therefore how much investigative burden and how much unwarranted suspicion the system generates. Publishing the number that lets someone calculate your false positive load is a real disclosure and the grade reflects it.

What is absent is any analysis of who absorbs that load. No breakdown by role, shift, unit or seniority has been published, and a model drawing on overrides, timing and time clock data will concentrate flags on the staff whose work is most interrupted and most irregular. There is also no governance statement covering model monitoring, drift or review.

No published position on responsibility, challenge or remedy, for a product whose output routinely becomes an accusation against a named clinician and frequently enters a disciplinary, licensure or law enforcement process. Nothing states who owns the determination, whether the clinician is informed that a model flagged them, what evidentiary weight the score is intended to carry, or how a wrong flag is withdrawn.

The published specificity figure makes this sharper rather than softer, because it establishes that wrong flags occur at a knowable rate and the vendor has therefore quantified the harm it declines to describe a remedy for.

Integration breadth is the product's structural advantage and it is specified rather than implied. Named sources include the electronic health record, automated dispensing cabinets, CII Safe, the narcotics vault, reverse distributor systems, retail pharmacy systems, wholesaler systems and the employee time clock.

Reaching outside clinical systems into the supply chain and workforce records is what allows reconciliation from purchase through to administration, and no competitor in this lane names a comparable span. The grade holds below the top band because specific electronic health record and dispensing cabinet vendors are not named, and nothing published describes the depth or method of each connection.

Nothing published describes where this product runs. No hosting model, cloud provider, region or residency commitment was located across two passes. The parent operates a documented enterprise cloud architecture and names its host and certification scope for products in other divisions, so the information plainly exists inside the company, and applying it to this product would be an assumption rather than a finding. A buyer evaluating a system that aggregates identified clinician behaviour from eight source systems cannot currently establish which jurisdiction that aggregate sits in.

Two passes located no price, no pricing page, no packaging tiers and no published basis of charge for the diversion module or the wider platform. The parent is publicly listed and reports segment revenue, which places this in the same position as every large vendor in this lane: a shareholder can size the business and a hospital cannot size the invoice.

Nothing was found describing whether the module is licensed per facility, per bed, per clinician monitored or per transaction, which matters more than usual here because the product's own value argument is built on transaction volume.

Coverage is defined by where medication actually moves rather than by care setting alone, which is the more useful cut for this problem. The product spans pharmacy, nursing and anaesthesia inside the hospital and reaches either side of it through wholesaler and reverse distributor data, retail pharmacy systems and the narcotics vault. It monitors non controlled medications alongside controlled ones, which most competitors do not attempt and which closes a real detection gap.

The published deployment evidence is concentrated in acute care inpatient hospitals, so post acute, long term care and ambulatory settings are asserted rather than demonstrated, which is what holds it below the top grade.