Etiometry

Split product and the split is real. The four risk indices are separately FDA cleared algorithms, patented, and the source of the company's name and differentiation. But the data aggregation and visualisation layer is a substantial product in its own right, and the company's own summary leads with consolidating ICU data before it mentions calculating risk indices. Strip out the indices and a hospital still has a vendor neutral device integration, waveform archive and quality improvement platform that many customers would still buy. That is the platform not algorithm test, and it lands at B rather than A. It does not fall to C, because unlike a pure data pipe the indices are cleared devices with published indications and are what the published evidence is about.

The most complete oversight disclosure in the category. The platform is repeatedly and explicitly positioned as not an active patient monitoring or alarm system, providing adjunctive information and informational flags, with instruction not to rely on it as the sole source of patient status information and an express statement that clinicians remain responsible for all assessment and treatment decisions. Nothing is auto actioned. Pathway automation highlights when HOSPITAL DEFINED criteria are met, so the institution rather than the vendor sets the trigger, which is the property that earned Affineon and Haystack credit. THE UNUSUAL PART, and the reason this reaches A: the company polices its own marketing against its own regulatory labelling, stating that its outcome studies do not establish predictive performance and do not expand the cleared indications. Vendors almost universally let favourable studies imply broader capability than their clearance covers. Doing the opposite in public is the strongest form of this disclosure the index has seen. Absent: per index sensitivity, specificity and false negative behaviour are not published on the site.

THE STRONGEST ON THIS AXIS IN THE CATEGORY AND THE DIRECT INVERSE OF THE EPIC SEPSIS MODEL, which scores D. Three things earn it. First, the mechanism is named precisely rather than gestured at: a Bayesian modelling approach combining mechanistic models of human physiology with techniques from aerospace navigation, which is a different lineage from the outcome trained classifiers used everywhere else in this lane. Second, the inputs are enumerated: heart rate, blood pressure, SpO2 and, where available, filling pressures, venous oxygen saturation, hemoglobin and blood gases. Third and decisively, THE QUANTITY BEING ESTIMATED IS DEFINED AGAINST PUBLISHED CLINICAL THRESHOLDS: hyperlactatemia as lactate above 4 mmol per litre, acidemia as arterial pH below 7.25, inadequate oxygen delivery against a configurable mixed venous oxygen saturation threshold. A clinician can state exactly what the index is estimating and can check it against a laboratory value. That is falsifiable from outside, which is the property this index found absent in Epic and in Isomorphic Labs. A public patents page supports it. Absent: model equations, calibration curves and per index operating characteristics, none of which appear on the site, though the last will be in the 510(k) summaries.

The deepest published evidence base in this category, and it is investigator led rather than vendor run. Verified citations include Gazit AZ et al., Critical Care Medicine 2025;53(7):e1355 to e1364, doi 10.1097/CCM.0000000000006682, a multicentre before and after study across three paediatric cardiac ICUs at tertiary academic children's hospitals reporting shorter vasoactive infusion duration, with cardiac arrest rate and weaning failure carried as secondary outcomes; Gaies M, Circulation: Cardiovascular Quality and Outcomes 2023 on readmissions; Clark MG et al., Journal of Pediatric Critical Care 2025, single centre retrospective, on ventilation duration; and work from ten hospitals including Boston Children's, Children's of Alabama, Children's National and St Louis Children's presented at the World Congress of Pediatric Cardiology and Cardiac Surgery. THE FRAMING IS THE MOST EPISTEMICALLY CAREFUL IN THIS INDEX: the company states on its own pages that results are observational, vary by site, do not establish causality, and do not establish predictive performance or expand the cleared indications, and its home page carries the line that the platform is intended to aid decision making and not to improve outcomes independently. Caveats held on the record: no randomised trial exists, before and after designs cannot exclude secular trend, the 36 percent length of stay figure traces to a Circulation supplement conference abstract rather than a full paper, the readmissions figure traces to a paper about causal inference methods, and the evidence is heavily concentrated in paediatric cardiac surgery so adult and general ICU performance is far less established.

Unusually candid about two things most vendors leave silent, which is what earns the B, while the governance around them is unpublished, which is what caps it. The company states that every data point ingested is retained indefinitely in its Quality Improvement Application with full fidelity waveforms archived separately, and separately states that data collected from hospitals globally drives continuous improvement of its risk algorithms, describing a corpus that has grown from 150 million to more than 200 million hours of patient physiologic signal. INDEFINITE RETENTION AND TRAINING USE ARE BOTH DISCLOSED OPENLY RATHER THAN BURIED, which is better than the silence this market usually offers, but no de identification method, customer opt out, contractual basis for the research corpus or retention limit was located. Architecture is a genuine counterweight: the platform runs on hospital installed virtual servers on hardened RHEL with no inbound internet traffic required, so processing sits inside the hospital environment.

HIPAA compliance is asserted, described as being built to HIPAA requirements for handling ePHI, but no business associate agreement posture is published: no statement of whether a BAA is executed as standard, on what terms, at what cost or with what subprocessor disclosure. A bare compliance assertion without BAA terms is the weak rung of this index's HIPAA ladder. Graded C rather than Not Rated because a claim is made and can be assessed, unlike vendors where nothing was located at all.

The best security disclosure in this category by a wide margin and among the strongest in the index. SOC 2 TYPE II WITH THE TYPE SPECIFIED, which is the question this index asks of every SOC 2 claim and which Affineon, RXNT, Supanote and others fail. Plus ISO 27001 certification, ISO 13485, MDSAP, and stated alignment with FDA Section 524B, the medical device cybersecurity provision, which is a specific and checkable claim rather than a generic assurance. Concrete technical controls are published rather than implied: TLS 1.3 in transit, access governed entirely through the hospital's own Active Directory with location based role restriction via AD groups, configurable idle reauthentication defaulting to 15 minutes, hardened RHEL servers, and no inbound internet traffic required. WORTH CREDITING SEPARATELY: the company correctly labels ISO 13485 as medical device quality management and ISO 27001 as information security rather than conflating the two, which is the exact framework confusion this index recorded against other vendors. The one gap is the absence of a public trust centre with a report request path.

The deepest regulatory footprint in this category, exceeding CLEW's. Four distinct cleared physiologic risk indices with population specific indications published per index: IDO2 for adults and paediatric patients above 2 kg, IVCO2 for invasively mechanically ventilated patients from 0 days to 12 years, plus HLA and ACD. Additional clearances cover the platform itself and persistent bedside display functionality, and in 2026 the Cardiogenic Shock Tool was cleared under K254066. International authorisations include EU MDR, Health Canada and MDSAP. INCONSISTENCY RECORDED, and it matters more here than it would elsewhere because regulatory depth IS this company's stated differentiation: the total clearance count is given as eight in a 2023 press release, as 10 in one answer on the current FAQ page and as 11 in another answer on the SAME page, and IDO2's first clearance is dated 2014 in the FAQ table while the company's own press release announces that first clearance in October 2016.

Two real mechanisms, one large unaddressed exposure. The mechanisms: the company states that it continuously monitors its own algorithm performance in production against each ICU's real world patient population so that drift is identified proactively, which is a direct answer to the institutional variability problem the Epic Sepsis Model record identifies and a different answer from CLEW's regulator bound change control plan; and it publishes population limits per index and refuses to let outcome studies imply broader capability. THE EXPOSURE IS SPECIFIC AND CONCRETE. IDO2 takes SpO2 as an input, and pulse oximetry is well documented to overestimate arterial oxygen saturation in patients with darker skin pigmentation, a measurement bias the FDA has itself examined. An index estimating INADEQUATE OXYGEN DELIVERY that ingests a biased oxygenation input can inherit that bias directly, and would do so in the direction of underestimating risk in exactly the patients already subject to unrecognised hypoxaemia. Nothing published addresses this. No subgroup calibration by skin pigmentation, race, sex or age appears anywhere. Publishing subgroup calibration for IDO2 would be the single highest value disclosure this vendor could make. The drift monitoring claim is also unaccompanied by methodology, threshold or any published report.

The deepest interoperability disclosure in the index. EHR side: Epic, Oracle Health and Meditech, with ADT and laboratory results over standard HL7 interfaces and medications, procedures and diagnoses through a dedicated FHIR application, plus SMART on FHIR single sign on so clinicians launch from inside the EHR. Device side, which nothing else in this lane approaches: bedside monitors from GE Healthcare, Philips, Draeger, Mindray, Nihon Kohden and Spacelabs, ventilators from Medtronic, Draeger, Hamilton Medical and Getinge, ECMO and Cardiohelp systems, VADs, CRRT, cerebral oximetry and Edwards HemoSphere hemodynamic monitors, reached through medical device integration platforms including Capsule, Bedcomm and IBUS, at five second numerics and waveforms up to 500 Hz. BIDIRECTIONAL, which is rare: derived risk indices can be pushed back out over HL7 and REST rather than only pulled in, and hospitals get CSV, MAT and JSON exports plus API and SDK access. One caution: a separate FAQ answer claims integration without requiring HL7 interfaces, which contradicts the detailed answers on the same page.

Publishes what almost no vendor in this index publishes: the actual shape of an implementation. Hospital installed virtual servers running the software, device and EHR integration setup, user authentication and access point configuration, with completion of that work defined as a named Technical Readiness gate before clinical rollout begins. Training is described in stages, super user sessions then unit end user training then help desk education before go live, with ongoing clinical meetings and continuous education afterwards. Architecture is hardened RHEL with no inbound internet traffic required, and access is governed through the hospital's own Active Directory, so data residency sits with the customer. Scale is stated at roughly 150 ICUs and 3,500 beds across 50 hospital partners. Two caveats recorded: the stated two month technical implementation is framed as a best case with no typical range given, and the ICU count is given as over 175 in one place on the FAQ and roughly 150 in another.

No pricing published at any level: no rate card, no unit of pricing such as per bed or per monitored patient day, no indicative band and no implementation or integration fee, despite the implementation itself being described in detail. Everything routes through a demo request. Worth recording separately because it cuts against the company's evidentiary strengths: the clinical research resources are gated behind a form that requires a work or hospital email address and explicitly rejects personal domains, so the published evidence a buyer would use to evaluate the product is not openly reachable. Gating clinical evidence behind lead capture is the same pattern this index documents in the analyst firms it competes with.

Coverage is broad and, more importantly, its limits are published per index rather than asserted generally. IDO2 is indicated for adults and paediatric patients above 2 kg, IVCO2 for invasively mechanically ventilated patients from 0 days to 12 years, with HLA and ACD carrying their own indicated populations. Settings span ICU, CICU, PICU and NICU plus step down and other high acuity inpatient units, and the platform is designed to follow a patient across acuity transitions rather than resetting at transfer. THE NEONATAL AND PAEDIATRIC DEPTH IS THE DECISIVE SPLIT IN THIS CATEGORY: CLEW is cleared for adults 18 and over only, while Etiometry originated in neonatal and paediatric cardiac critical care. Deep speciality grounding in cardiac surgical critical care is named down to condition level, covering STEMI and NSTEMI, acute and decompensated heart failure, right, left and biventricular dysfunction, and the perioperative period around CABG, valve surgery, transplant and congenital repair. Caveat carried from the evidence axis: coverage is broad but the outcome evidence is concentrated in paediatric cardiac academic centres.