DoseMeRx

The dose is produced by a learned component, and that component is the product. DoseMeRx estimates each patient's pharmacokinetic parameters by Bayesian updating: a published population model supplies the prior, and the patient's own measured drug concentrations, and genotype where available, update it to an individual estimate from which the regimen is computed.

Everything else on the platform, the fit display, the renal trajectory, the guideline comparison and the dosing matrix, presents that estimate to a clinician; remove it and nothing is left to dose from. The mechanism is statistical inference with a long clinical pedigree rather than a neural network. DoseMe calls itself a Bayesian dosing platform throughout its material rather than reaching for an artificial intelligence label, which describes the method accurately.

What the clinician is shown alongside the answer is the oversight design. The interface displays how well the model fits this patient's observed levels rather than presenting a dose as a bare number, tracks serum creatinine on the same page so renal trajectory is visible while the decision is made, and compares the individualized result against guideline recommendations.

It presents an Alternative Dosing Matrix of viable options across dose amount and interval, any of which can be selected, rather than a single recommendation to accept or reject. The system never acts on its own: every dose is a clinician's choice from a displayed option space, with a fit diagnostic and a guideline reference beside it, which is meaningful oversight in practical form. The remaining gap is a published figure for how often clinicians adopt, modify or reject the recommendation.

What produces the dose is named. The engine is Bayesian estimation over published population pharmacokinetic models, and DoseMe names them with their sources: for vancomycin, Buelga et al. (2005), Goti et al. (2018) in one and two compartment and hemodialysis forms, Sabourenkov et al. (2019) for obese adults, Lamarre et al. (2000) for pediatric and adolescent patients and Frymoyer et al. (2014) for preterm neonates, each with its structure, covariates and derivation cohort described.

The platform also exposes model fit at the point of care, so a clinician can see whether the model describes this patient's observed levels before acting. What is missing is version and update discipline: the drug models page shows each model's development stage, but nothing says how a customer learns that a live model has changed, no calibration data is published for the library, and the genotype input has no published description of which gene and drug pairs are supported or how genotype is weighted against measured concentrations.

The chain is structurally short and its terms are published. The dose models are published population pharmacokinetic models, named with their source studies on DoseMe's drug model pages, run on DoseMe's own platform hosted in Microsoft Azure, in Illinois for United States data and in EU data centers for European data.

The data security page states that customers own the patient data they enter, and discloses precisely what DoseMe takes in return: a license back over deidentified data to deliver, monitor and improve the service and improve the models in use, with reports and research exports as the data's owner determines; the HIPAA notice adds that deidentified pooled data may be shared with third parties for research unless a patient restricts it.

Secondary use is disclosed and bounded as a license from the owner. DoseMe also requires storage facilities to be audited against ISO 27001, SOC 1, PCI DSS and HIPAA security standards. No subprocessor list names every party that touches the data, including the third parties that receive research data.

DoseMe lists peer reviewed studies of its software on a clinical evidence page, most written by hospital clinicians rather than the company. Among them: a 2023 study in the Journal of Pharmacy Technology associating DoseMeRx vancomycin dosing with lower acute kidney injury incidence and better target attainment than first order AUC calculation in patients also on piperacillin-tazobactam; a 2024 Therapeutic Drug Monitoring report from a 500 bed hospital showing lower kidney injury rates, fewer serum samples and cost savings after switching to Bayesian AUC monitoring; a 2021 prospective and retrospective quasi experimental study in the British Journal of Clinical Pharmacology showing more patients reaching target AUC; and pediatric cystic fibrosis tobramycin studies in the Journal of Antimicrobial Chemotherapy (2022) and Clinical Pharmacokinetics (2015, 2018).

An independent 2018 comparison in Pharmacotherapy included DoseMe among five programs. This published research is largely retrospective and single site. No prospective multi site study backs the outcome claims, and reduced kidney injury is the one that carries the clinical argument.

The company publishes specific positions on training use and on deidentification. Training use: its data security page states that patient data entered by customers is owned by those customers, and where applicable by the patient, and that DoseMe licenses back de identified data to deliver, monitor and improve the service and to improve the models in use, with reports and research exports as determined by the data's owner.

De identification: its HIPAA notice says de identified data is used whenever possible to maintain and improve the service, that de identified and pooled data may be shared with third parties for research, and that a patient can restrict that use through a signed HIPAA consent form. What is not published is a retention window for patient data or for the de identified derivatives. Supporting controls are described: encryption, firewalls, full user audit logging and encrypted mobile caching. No safety engineering is published: nothing describes guardrails on a dose recommendation beyond the fit display a clinician sees, or how a safety event is handled.

DoseMe publishes a HIPAA page carrying a Notice of Health Information Practices for DoseMe Operations Inc., describing how protected health information is collected and used, what is not done with it, that de identified and pooled data may be used and shared for research, and that a patient can restrict that use through a signed HIPAA consent form, with a named contact for HIPAA inquiries. The same page states where United States data is stored. Two limits.

The notice is written in the voice of a provider of health information practices and does not use the term business associate, so the contractual role DoseMe takes is not stated. And the business associate agreement itself is not published, so its breach notification timelines, subcontractor flow down terms and data return obligations on termination are not public. A separate page addresses the EU General Data Protection Regulation for European customers.

DoseMe holds HITRUST certification and states that it is the only Bayesian dosing platform to do so, a claim that can be checked. Its data security material describes multi layered controls, including physical security, firewalls, encryption, full user audit logging and encrypted mobile caching, and commits to storing data only in facilities audited against ISO 27001, SOC 1, PCI DSS and HIPAA security standards; those facility audits belong to the hosting provider rather than to DoseMe, and are the right obligation to pass down rather than an attestation of DoseMe's own.

No HITRUST date, tier or scope is stated, no service organization controls report of DoseMe's own is published, and no trust center, penetration testing cadence or vulnerability disclosure program is described.

The software makes no device claim and is scoped to match: DoseMeRx presents dose options and a model fit display to a pharmacist or physician who chooses the dose. No United States clearance is claimed. The open question is broader than for a United States only vendor, because the product is sold in the United States, Europe, Asia Pacific, Africa and South America, with a named hospital customer in Portugal and a published GDPR page, and no CE mark, UK conformity marking, Australian Therapeutic Goods Administration listing or other national designation is published, nor any statement of why none is needed. The drug models page says models are implemented following regulatory guidelines for Bayesian dosing, without naming them.

DoseMe publishes a governance process for its models. Its drug models page says a team of bioinformaticians, scientists and clinical pharmacists selects clinically validated pharmacokinetic and pharmacodynamic models, that each goes through a validation process following quality assurance practice, ISO standards and regulatory guidelines for implementing Bayesian dosing, including an independent review by its Clinical Advisory Board, and it shows each model's stage from evaluation through review, development and validation to market.

The Clinical Advisory Board is a named body, chaired by a named clinical pharmacologist and described by the company as advising on clinical excellence and governance. Evaluation is the gap: no performance reporting by age, sex, race, ethnicity, renal function band or body habitus and no bias assessment is published. The population specific models the company offers, for obese, dialysis, pediatric and neonatal patients, show it treats covariate fit as a design question.

One mechanism is well placed. The platform exposes model fit to the user at the point of care, so a clinician can see whether the model describes this patient's data before acting on what it predicts. That is a per patient confidence signal at the moment of decision, more useful than an aggregate accuracy figure because a dosing model can be good overall and wrong for the patient in front of you, and it gives the clinician a defensible basis for overriding the recommendation.

The models behind each drug are named with their source studies, so a pharmacist can check which published cohort underlies a prior. Nothing stands behind the output: no warranty, indemnity or remediation commitment is published, no calibration data or error rate is published for the library, and nothing describes how genotype and measured concentration are reconciled when they disagree.

Integration with major hospital systems including Epic and Cerner is long established and predates the current ownership, and a customer has announced an integration with Oracle. The most consequential integration runs through another vendor's product: in October 2025 Inovalon announced a collaboration bringing DoseMeRx dosing into its pharmacy surveillance workflows, so the software reaches clinicians inside a surveillance platform as well as directly. No FHIR conformance statement, public API or electronic health record marketplace certification is published, and the direction and depth of each integration are not documented.

Residency is stated precisely. The HIPAA privacy statement says United States healthcare data is stored in Microsoft Azure in an Illinois based facility, the data security page says DoseMe stores data in the jurisdiction in which it is owned and that European data is held in Microsoft Azure data centers within the EU, and the privacy policy commits to on shore storage and sets out what happens if an overseas transfer is needed, binding overseas recipients to the Australian Privacy Principles.

The platform is cloud delivered with no on premise option described. Isolation is left open: nothing is published about how one customer's data is separated from another's in the shared Azure environment. A privacy office can otherwise answer its residency question from the website.

The price is undisclosed and two concrete commercial terms are published. A fourteen day free trial covers the full platform including all drug packages, so an evaluating pharmacy is not restricted to a limited module set. A partnership with the American Society of Health System Pharmacists aimed at small and rural hospitals bundles that society's competency management subscription with unlimited use of the platform for up to ten users at a single site, stated to be open to non members.

That publishes a packaging unit, the site and the user count, so how the product is sold and what drives the cost are visible before contact. No price, contract term or renewal structure is disclosed for either the bundle or the standard product, and it is not stated whether the trial converts automatically.

Where the product is validated to operate is named and supported. The drug models page marks each drug's models as adult, pediatric or neonatal, and the vancomycin page names population specific models with their source studies: adult one and two compartment models, a hemodialysis model, an obese adult model developed on 2,993 courses in patients with BMI over 30, a pediatric and adolescent model, and a preterm and neonate model supporting dosing from 22 weeks gestation.

Peer reviewed studies support use in obese adults, children with cystic fibrosis and resource limited hospitals. Settings are named too: acute hospitals, infusion service centers, a program of all inclusive care for the elderly, and small and rural hospitals reached through a bundle with the American Society of Health System Pharmacists. More than a thousand clinicians across more than a hundred locations use it in the United States, Europe, Asia Pacific, Africa and South America. The function itself stays narrow: high risk medications with therapeutic drug monitoring.