InsightRX

The computational core is quantitative pharmacology rather than machine learning. Nova runs published population pharmacokinetic models and performs maximum a posteriori Bayesian fitting against a patient's measured drug concentrations, which is statistical inference with a long clinical pedigree and no learned component. Real machine learning sits above that layer and is not decorative.

The company built a feature that assesses a patient, a model and the available measured levels to decide whether fitting with flattened priors will improve the prediction, and it has published a machine learning approach to pharmacokinetic model selection and averaging in a peer reviewed pharmacometrics journal. Model refinement is described as a human assisted process combining pharmacology expertise with machine learning techniques.

Graded B and the candour is credited: the company leads with Bayesian forecasting and quantitative pharmacology in its own copy rather than reaching for an artificial intelligence label it could easily have claimed, which places it alongside the other honest non inflators in this index.

A pharmacist sits between the model and the patient by design, since the output is a proposed regimen entered into a review workflow rather than an order. Two disclosed design decisions are better than the category norm and both concern what happens when the inputs are bad.

Nova flags missing or stale data explicitly within the interface rather than silently computing on it, and pharmacists retain full manual data entry so that a broken feed or an electronic health record outage degrades the workflow rather than blocking patient care. That directly addresses the substrate quality exposure this index has documented elsewhere, where a model consumes whatever arrives without signalling that it is unreliable.

Held at B because no threshold or confidence bound is published at which the software declines to make a recommendation, and no figure is published for how often pharmacists accept the proposed regimen unchanged, which is the acceptance rate equivalent this category still lacks.

This is the most inspectable model stack graded in this index so far and the grade reflects a difference in kind rather than degree. The pharmacokinetic models Nova runs are named published models from the literature, including Thomson, Goti, Carreno and Hughes, so a pharmacist can read the equations and the cohorts they were derived from without the vendor's permission.

The company publishes which model it sets as default for which patients and the evidence behind that choice, having analysed more than 168,000 treatment courses to conclude that a modified Thomson model performed better in patients below a body mass index of 40 and changed its default accordingly. Its machine learning method for model selection and averaging is published in a peer reviewed pharmacometrics journal with the workflow described.

The comparison worth drawing is with the opioid risk score in this same category, where the model is a ten input logistic regression whose coefficients are unpublished. Complexity and opacity are independent, and this vendor proves it in the direction that helps the clinician.

The evidence is real, partly independent, and reported here including the part that does not flatter the vendor. An independent group publishing in Pharmacotherapy compared five Bayesian dose optimising programs in critically ill patients and found InsightRX the most adaptable and best supported of the five while ranking a competitor as the least biased, so the one genuinely independent head to head in this segment did not put this vendor first on accuracy.

A separate group in the Netherlands published a prospective validation in intensive care patients reporting bias and imprecision within a clinically acceptable range. Company authored work with employment and equity disclosed reports that switching the default model improved area under the curve target attainment from 46.5 to 51.1 percent and cut median time to target from 47.8 to 28.2 hours, while severe acute kidney injury rose from 1.75 to 2.03 percent with a number needed to treat of 358.

A paediatric intensive care randomised pilot using Nova is registered. Held at B because the outcome work is company authored and the strongest independent comparison ranked a competitor ahead on the accuracy measure that matters most.

Data reaches the platform through standards based integration rather than bulk extraction, with automated import over SMART on FHIR limited to the patient data the dosing calculation requires, which is a narrower footprint than most vendors in this category take. HITRUST certification and stated GDPR compliance cover the handling controls, and the interface behaviour around stale and missing data is a genuine safety property rather than a compliance one.

The open question, and it is the one to put to the vendor, concerns secondary use. The company describes its pharmacokinetic models as continuously improved using real world clinical data and sells a separate analytics product built on aggregated dosing data, so a customer should establish exactly what governs the use of its patients' concentration and outcome data for model development, whether that use is opt in, and whether the resulting models are then sold back to the wider market.

HIPAA compliance is asserted on a dedicated security and regulatory page rather than in marketing copy, and it sits inside a control set that includes HITRUST certification and GDPR compliance, which makes it more substantial than a bare claim since the HITRUST framework incorporates the HIPAA security rule. Held below the top grade for the usual reason in this index: the instrument itself is absent.

Retrieval located no business associate agreement text, no tier structure, no execution path and no subprocessor list. A buyer sees a credible assertion backed by an audited framework, which is a good position, and still has to request the agreement to learn what it actually says about breach notification, subcontractor flow down and data return on termination.

The posture is strong and shaped differently from most vendors here, because it is built around medical device software standards rather than only enterprise security attestations. A dedicated security and regulatory page names HITRUST certification alongside conformance to ISO 13485 for quality management, IEC 62304 for medical device software lifecycle processes, ISO 14971 for risk management and 21 CFR Part 11 for electronic records.

Very few vendors in this index claim the software lifecycle and risk management standards at all, and for software that computes a dose they are arguably more informative than a general security attestation, because they speak to how the code is developed and how hazards are analysed.

Held at B rather than A because no SOC 2 report of either type was located, there is no trust centre with retrievable documents, and no penetration testing cadence or vulnerability disclosure programme is published.

The precise position is published and it is weaker than the marketing around it suggests, so both are recorded. The company's security and regulatory page states that Nova is self certified as a Class I medical device under the European Medical Devices Directive, carrying a CE mark valid in the European Union and the United Kingdom, and that page deserves credit for stating the route exactly.

Self certification at Class I is the lightest regulatory pathway available in Europe and involves no notified body review of the software. A 2021 company announcement described the same milestone as having cleared requirements as a medical device with the European Medicines Agency and United Kingdom regulatory authorities, which overstates it, since the European Medicines Agency does not regulate medical devices and no regulator assessed the product. Two live questions for a buyer.

The Medical Devices Directive has been superseded by the Medical Device Regulation, whose classification rules for software that informs treatment decisions are materially stricter, so current European status should be confirmed directly. And there is no United States clearance of any kind, meaning the same software that is a regulated device in Europe is unregulated decision support in America.

This is the strongest governance posture in the lane so far and it arrives in an unexpected form. The company published an analysis of more than 168,000 treatment courses showing that model accuracy differed by body mass index band, then changed its default model for one of those bands and said so publicly.

That is stratified performance analysis acted upon and disclosed, which is the substance of what this axis asks for even though the stratifying variable is physiological rather than demographic. It also published a harm signal alongside a benefit, reporting that improved target attainment came with severe acute kidney injury rising from 1.75 to 2.03 percent and giving the number needed to treat, when omitting the second half would have been easy and unremarkable.

Held at B because no analysis by age, sex, race, ethnicity or socioeconomic factors was located, which matters because population pharmacokinetic models are typically derived from convenience cohorts whose composition determines whose physiology the priors describe.

Integration is certified rather than asserted, on two major marketplaces and by a published standard. The applications are fully certified in the Epic Showroom with automatic patient data import, certified in the Oracle Cloud Marketplace for Oracle Health, newly available for direct integration with MEDITECH Expanse, and the underlying mechanism is SMART on FHIR rather than a proprietary interface, with the vendor's implementation team validating data flow before go live.

There is a second and unusual distribution path: the precision vancomycin dosing tool inside Premier's TheraDoc clinical surveillance platform is InsightRX, so the software also reaches clinicians through an incumbent competitor's product. Named deployments span Sentara Healthcare in the United States and Amsterdam University Medical Centers, Máxima Medical Center and University Hospital Würzburg in Europe.

Nova is described consistently as a cloud based platform integrating into hospital systems, and it operates on both sides of the Atlantic, with a CE mark covering Austria, Belgium, France, Germany, the Netherlands and the United Kingdom and named European academic hospital deployments. Stated GDPR compliance implies a European data handling position.

What is absent is the specificity a privacy office needs: retrieval located no named hosting provider, no cloud region, no data residency commitment, no statement of whether European patient data remains in Europe, and no on premise or private cloud option. For a vendor operating clinically in six European countries the residency question is concrete, and answering it publicly would be straightforward.

Pricing is undisclosed across the provider pages, the platform pages and the press record. No price, no unit, no pricing basis, no contract shape and no implementation fee guidance were located, and the company does not indicate whether the commercial unit is the institution, the bed, the drug module or the dosing calculation, which for a product sold as a library of more than a hundred drug modules is a material difference.

Support and onboarding are described in detail, including hands on implementation, live training, ongoing clinical pharmacist guidance and around the clock technical support, so the service wrapper is well specified while its cost is not. Publishing the pricing basis alone would move this grade.

Breadth is genuine on three dimensions. The library covers more than a hundred drugs and therapeutic areas spanning infectious disease, oncology and transplant, rather than the single antibiotic that dominates this segment. Age coverage extends to paediatrics, with a registered randomised pilot in a paediatric intensive care unit dosing vancomycin and amikacin. Geographic coverage spans the United States and six European countries with named academic hospital users.

Held at B for a gap between the claimed library and the visible one: vancomycin accounts for almost all of the published evidence and most of the customer commentary, so a buyer adopting the platform for oncology or transplant dosing is relying on modules with a far thinner public track record than the headline count implies. Ask which specific modules the institution will use and what validation exists for each.