EndoTool

The algorithm is the most technically substantial in the glycemic segment and it is still not machine learning. Independent literature describes model predictive control with a nonlinear dosing equation that is individualised to the patient and then optimised over time based on the glucose responses to doses already administered, using diabetes type, height, weight, creatinine, age and sex as inputs.

Model predictive control optimises against a predicted trajectory rather than reacting to the current error, which is a genuine step beyond the proportional control its principal competitor uses, and the individualisation is real rather than nominal. The same peer reviewed reviews nonetheless group this product with rule based systems carrying static decision rules rather than adaptive learning models.

Graded C rather than lower because the mathematics does meaningful patient specific work, and because the vendor describes it accurately as a patented patient specific algorithm rather than claiming it learns.

A provider orders the protocol, the software computes each dose, and a nurse administers it, so no recommendation reaches a patient without a licensed clinician acting. Coverage of the transition from intravenous to subcutaneous insulin is worth noting on this axis specifically, because that transition is one of the most error prone moments in inpatient glycemic management, where an infusion with minutes of half life is replaced by a subcutaneous dose lasting hours and the consequences of getting it wrong are delayed rather than immediate.

A system that computes both sides of that handover removes a manual calculation at exactly the point where manual calculation most often fails. Held at B because no override rate, acceptance figure or published account of the fallback workflow during system unavailability was located.

Two partial routes to the mechanism exist. The algorithm is patented, and patents are public documents, so a determined reader can reach the method in a way most vendors do not permit even in principle. And independent clinical reviews describe the control strategy, the nonlinear form of the dosing equation and the covariates it consumes, which is enough for a clinician to understand the shape of the computation.

What the vendor itself publishes is a functional description with no equation, no parameters, no thresholds, no calibration data and no statement of the conditions under which the algorithm should not be relied upon. Graded C on the same basis as its competitor: the information is obtainable and the vendor is not the one providing it.

The published base is real but thinner than its principal competitor's, and the most useful entries are the unflattering ones. An independent evaluation in a burn intensive care unit reported time in target range improving from 41 to 47 percent with no significant reduction in hypoglycaemia at any of the thresholds examined, which is a modest result honestly reported and is exactly the kind of finding this index looks for.

The product also appears as the comparator in studies published by its main competitor, including a retrospective single centre comparison that changed the target glucose range at the same time as the software and therefore cannot separate the two effects, and a claimed head to head advantage in severe hypoglycaemia events that is asserted by the competitor rather than published jointly. A buyer should treat every head to head figure in this segment with care, since both available comparisons were produced by one of the two parties.

The data required is narrow and specific: glucose values, insulin administered, and a small set of patient characteristics including renal function and body measurements. The acquisition changes the stewardship question rather than the footprint, because the new parent operates an outpatient remote monitoring platform holding continuous glucose and device data, and has stated an intention to build a single integrated hospital to home solution.

Joining an inpatient dosing record to an outpatient monitoring record produces a longitudinal diabetes dataset neither product held alone. Retrieval located no retention schedule, no de identification statement and no description of whether or how inpatient data will flow into the combined platform, and that is the question a hospital should put in writing before the integration proceeds.

Two retrieval passes located no HIPAA statement, no business associate agreement terms, no execution path and no compliance page for the product or its immediate corporate parent. The acquisition adds a specific and answerable question rather than a general one: when a product changes corporate hands, the business associate agreements governing existing hospital customers either transfer, are novated or are replaced, and which of those happened is a matter customers are entitled to know.

Nothing published addresses it. Hospitals running the product should confirm which entity is now their business associate, since the clinical literature alone refers to the developer under two different company names.

Retrieval located no service organisation controls report, no HITRUST certification, no ISO 27001, no medical device software lifecycle or risk management standard, no trust centre, no penetration testing cadence and no vulnerability disclosure policy.

As with its competitor the omission of the device software standards is the more notable half, since a quality system and a documented hazard analysis must exist to sustain the clearance and naming them would communicate more to a hospital than a general security attestation.

The recent change of ownership makes publication more useful than usual, because a customer currently has no public basis for judging whether the security and quality posture that held under the previous owner continues unchanged.

Cleared by the Food and Drug Administration for intravenous insulin dosing, for the transition from intravenous to subcutaneous therapy, and for subcutaneous dosing, in both adult and paediatric hospitalised patients, which makes it one of only a small number of computerised glycemic management systems holding clearance across that range. Independent literature consistently lists it among the cleared systems available in the United States.

As with its competitor the record demonstrates the regulatory boundary that defines this category: software that calculates an insulin dose is treated as a device requiring clearance, while software that alerts a clinician about a medication risk elsewhere in this same category operates under the statutory exclusion for non device clinical decision support. A buyer should confirm the current clearance holder following the change of ownership, since a device clearance attaches to a specific establishment.

Retrieval located no performance reporting by any patient characteristic, no bias assessment and no post deployment monitoring statement. The exposure here is somewhat different from its competitor's because this algorithm explicitly consumes creatinine, weight, height, age and sex, so the covariates that drive individualisation are known and the question of how performance varies across their range is directly answerable from data the vendor already holds.

Renal impairment is the case that matters most, since insulin clearance falls as kidney function declines and the patients most likely to be on an insulin infusion are disproportionately those with impaired renal function. Performance stratified by renal function band is the specific disclosure to request.

The product is described as integrating directly with major electronic health record systems, which is a precondition rather than a differentiator for inpatient dosing software, since the algorithm needs glucose results and administered doses in near real time to function at all.

The acquisition adds an outpatient dimension, since the new parent's platform is built on unified device integration and record connectivity across continuous glucose monitors and insulin delivery devices, and the stated ambition is a single solution spanning hospital and home.

Held at B because no named marketplace certification, FHIR conformance statement or public interface documentation was located, and because the hospital to home integration is announced intent rather than a shipped capability a buyer can evaluate today.

The system is delivered as networked hospital software with electronic health record integration, and public material does not describe whether it is hosted by the vendor, deployed inside the hospital's environment, or offered both ways, which for critical care dosing software is a material architectural question rather than a preference. Retrieval located no named hosting provider, no region, no residency commitment and no description of degraded mode behaviour when connectivity fails.

The ownership change compounds the gap, since a hospital cannot tell from public material whether the hosting arrangements that applied under the previous owner persist or whether migration to the acquirer's infrastructure is planned.

No price, unit, pricing basis, contract shape or implementation fee was located for the product before or after the acquisition. The acquisition itself introduces the commercial question a customer should ask now rather than later: the acquirer has stated that both product lines will continue to be supported in the near term and has described an intention to build an integrated hospital to home platform, so establish contractually what continued support means, how long it is committed for, and whether future pricing assumes purchase of the outpatient platform alongside the inpatient dosing system. Bundling pressure after an acquisition is ordinary and is easier to negotiate before renewal than after.

Within inpatient glycemic management the coverage is complete: intravenous dosing for critical care, the intravenous to subcutaneous transition, and subcutaneous dosing on the ward, across adult and paediatric patients. Independent evaluation exists in a burn intensive care unit, which is among the harder glycemic environments given the metabolic derangement involved.

The acquisition extends the theoretical reach across the discharge boundary into outpatient remote monitoring, though that is announced rather than delivered. Held at B because the function remains one drug in one disease, no use outside the United States was located, and the named install base is not published in the way its principal competitor's is, so the deployment scale cannot be compared.