Sickbay

AI is claimed and no learned component can be found in the product. The claims are the company's own: the patient monitoring page offers "AI-informed analytics and workflow tools", a news post says Sickbay uses "machine learning AI" to capture high resolution signals at UAB Medicine, and a blog post lists "Predictive Waveform Analytics" that detect compound physiologic patterns before single parameter limits are breached. The product pages describe something else.

Virtual Ops ranks patients by logic the clinicians configure, and the company states that the software generates no clinical interpretations or treatment recommendations. Risk indicators, thresholds and alarms are set by the hospital, per signal and per patient, and Sickbay Analytics gives hospital teams tools to build and deploy their own algorithms.

The predictive models published on Sickbay data, such as a single ventricle deterioration index developed at Texas Children's and validated at three other children's hospitals, are research teams' work and are not sold as a Sickbay product. A user quote on the home page names a "Decomp Score" with no method. Without the claim, Sickbay is a vendor neutral platform that captures, stores and displays bedside device data alongside EHR labs, medications and admissions.

Nothing in Sickbay acts on a patient. The 2015 clearance covers remote display and review of device data, waveforms and alarms to aid clinical decisions, and warns that the platform supplements the hospital's device monitoring and must not be the sole source of alarms. Virtual Ops ranks patients by logic the clinicians configure, and the company states that all decisions rest with qualified clinical staff. Thresholds and alarms can be left at defaults or set per signal and per patient.

One part is missing: what happens when a configured indicator misses a deteriorating patient or floods a unit with alerts. No shipped default indicator set, escalation path or review of misses is described, and the blog's predictive analytics claim is not tied to any described control.

The architecture is described in general terms, and the analytics the company calls AI informed and predictive are not explained. The 2015 510(k) summary describes software on the hospital's servers that routes and stores device data, stores EHR labs, medications and admissions received in HL7, and redisplays them. The company attributes the platform to a grid computing architecture from its chief product and innovation officer's research.

Virtual Ops outputs come from logic the clinicians configure. No method, model, version or training data is published for any predictive output, and no release or versioning practice is published for the platform since its 2015 clearance.

The chain is structurally short. The company says Sickbay is hosted only on the hospital's own network or cloud under the hospital's security standards, so patient data does not pass to the company by design, and no outside model sits between device data and a displayed indicator, which the clinicians configure. Two parts are left open.

The partner page describes Healthcare Intelligence Partners, a portfolio of clinical analytics, AI and workflow solutions powered by Sickbay data, and OEM partners who embed Sickbay's aggregation and storage in their own devices and platforms; none is named. Hospitals can also export data into AI pipelines and cloud tools of their choosing.

Named customers and outcome figures with no method. The company says Houston Methodist saw a 37 percent reduction in Code Blue rates after implementing Sickbay in its ICU, with no period, denominator or source, and the Virtual Ops page lists reduced mortality, length of stay and Code Blues with no figures.

The peer reviewed paper the company promotes, Mullen 2024 (J Nurs Care Qual, two company authors), reports alarms per bed per day falling from 177.69 to 96.94 and time in alarm from 7.52 to 2.83 percent over ten years on one pediatric progressive care unit at Texas Children's; its abstract credits alarm settings in the patient monitoring system and alarm champions, and does not name Sickbay.

Research use is broad: a 2025 systematic review counted 41 pediatric papers using Sickbay or Etiometry data, and the company cites over 100 publications, posters and presentations, which is scale of use rather than evidence of benefit. Named sites include Houston Methodist, Texas Children's, UAB Medicine and Michigan Medicine (rolled out in 2023).

General assurances that leave the questions open. The company says Sickbay is hosted only on the hospital's own network or cloud under the hospital's security standards, which keeps the archive in the customer's environment. The privacy notice of 12 May 2026 states that it does not cover services involving protected health information, and no document for the service addresses retention, de-identification or secondary use.

The 2015 510(k) summary says physiological data is stored on disk until deleted, so retention is left to the hospital. Nothing says what reaches the partner analytics and AI solutions the company describes, or whether the company uses customer data for its own development. No safety engineering is published, meaning how an analytics output is checked and how a safety event is handled.

Compliance is claimed without the document behind it. A company blog post says Sickbay "prioritizes data security and compliance with regulations, such as HIPAA" by deploying on the health system's own network or cloud. No business associate status is stated, and the one privacy notice published, dated 12 May 2026, excludes services that involve protected health information. Because the platform runs in the hospital's environment, whether the company acts as a business associate turns on its support, hosting and partner access, none of which is described.

Controls are asserted with nothing independent behind them. The company says Sickbay is hosted only on the hospital's network or cloud using the hospital's security protocols, and its cybersecurity post describes threats in general terms. No SOC 2, HITRUST or ISO/IEC 27001 attestation, penetration test, vulnerability disclosure route or trust center is published.

The 2015 clearance predates FDA's requirement that submissions for networked devices include cybersecurity documentation, so no review of that kind sits behind it. The platform keeps continuous waveform data for every monitored patient.

The platform holds a clearance; the analytics the company markets as AI informed and predictive carry neither a clearance nor a stated position. K143304 (26 May 2015, Class II, product code MWI) clears the Sickbay Clinical Platform, with a Patient Monitoring app, Patient Alarm Data Monitor and Alarm Analytics Dashboard, to route and store device data to the EHR and to display physiologic data, waveforms and alarms for remote consultation and review in a hospital, as a supplement to device monitoring and not the sole source of alarms.

The site calls the platform FDA cleared without the number, date or indication. The company states that Virtual Ops is not a device and sits under enforcement discretion as a command center visualization interface. The predictive waveform analytics, described as detecting deterioration patterns before alarm limits are breached, are not labeled cleared and have no position. FDA's database lists no later clearance for the company.

Policy language, no evaluation. The company publishes a post on data governance (FAIR principles, provenance, audit trails and data ownership agreements), a webinar on ethical AI deployment with Penn Medicine, and a white paper on trusted AI, and the Analytics page says hospitals can study their monitoring data to reduce potential bias in care. None of this evaluates a Sickbay output.

No capture completeness by device type or unit, and no performance of configured indicators across patient groups, is published. For a data layer that other models are built on, the governance question is what data goes missing and for whom, and what governs algorithms a hospital or partner deploys through the Analytics tools.

Mechanisms exist and nothing stands behind the output. Indicators and thresholds are configured by the hospital, so a clinician asking why an alert fired can read the configuration, and waveforms and events are kept and can be annotated and shared; a company executive says clients have shared the record with plaintiffs to resolve claims. The 2015 clearance warns against relying on Sickbay as the sole source of alarms. No warranty, indemnity or remediation commitment is published, and no error rate is published for the Alarm Analytics Dashboard or any analytics output.

Standards in both directions with named sites behind them, and no EHR named. The 2015 510(k) summary describes EHR labs, medications and admissions received in HL7 and stored for redisplay, and device data routed to the EHR; the company describes HL7 interfaces pushing cleansed telemetry data into the EHR for documentation and billing. No EHR vendor, FHIR interface or marketplace listing is named. Named sites include Houston Methodist, Texas Children's, UAB Medicine and Michigan Medicine. On the device side, the platform takes data from monitors, ventilators and other bedside devices across manufacturers.

Options and residency are stated. The company says Sickbay is hosted only on the hospital's own network or cloud under the hospital's security standards, and the 2015 510(k) summary describes the software on the hospital's servers with data stored on disk until deleted. Viewing runs in a browser on laptops, phones, tablets and wallboards, on site or in remote command centers, and the company states that monitoring and virtual care can be upgraded in less than 90 days. What is open is the processing path: whether the company or its partners reach the environment for support, analytics or updates, and from where.

No price is published, and part of the posture is discoverable. The company says the platform supports unlimited users and runs on the hospital's own infrastructure, sells through reseller partners that offer managed services, and markets savings on bedside integration hardware and remote nursing station costs. Nothing states the unit of charge (monitored beds, device interfaces or data volume), implementation fees, or who pays for connectivity hardware.

Coverage is named with validation behind part of it. The 2015 clearance covers hospital use, and its summary names ICU and emergency patients. The company now markets ICU, telemetry and virtual command center use, emergency departments, neonatal and pediatric care, and a rural health offering, and its post on the Intel Capital partnership mentions at-home monitoring, which the clearance's hospital setting does not reach. Published research on Sickbay data concentrates in pediatric cardiac intensive care. No age, population or unit exclusions are published.