VirtuSense VSTOne

The model is the engine of the product's lead use, and the device now carries other value too. Fall and pressure injury monitoring depend on it entirely: a LiDAR based sensor produces three dimensional point clouds, and on device AI reads subtle postural shifts that precede a patient getting up, which the company says gives 31 to 65 seconds of warning before an unassisted bed or chair exit, and recognizes when a patient has not moved for a set period.

Remove that and a depth sensor alone does nothing a nurse can use. But the same device is sold for virtual nursing and virtual sitting, with an optical pan, tilt and zoom camera streaming 4K video (up to 24 streams to one sitter), and for violent patient notifications driven by a flag in the record, so a hospital would still have a working virtual care device without the model.

An AI patient assistant (in beta) and location tracking with hand hygiene monitoring (in development) are listed as well. B because the monitoring capability the company leads with does not exist without the model.

The home page says VSTOne autonomously monitors for fall risk and pressure injury "24/7, without wearables, without cameras, without false alarms", and the hospital page repeats "without cameras, wearables, or false alarms" beside "Zero Alarm Fatigue". That is an absolute claim that the system's alerts are never wrong, with the product as the subject, and the company's own product page contradicts it by saying VSTOne eliminates more than 95 percent of false alarms. A unit that believes the claim has no reason to measure what the alerts get wrong.

The oversight structure behind the claim is otherwise described. The system notifies rather than acts: an alert reaches the assigned nurse's device, nurse call and room lights, and the company describes routing alerts to other staff when a nurse is already in the room with a patient flagged as violent. Monitoring can run fully automated, as a hybrid with a human virtual sitter, or as full virtual sitting. What is not published is the threshold at which an alert fires, how sensitivity is set by patient, and how often the system stays silent before a fall.

The approach is described specifically. The sensor stack is listed as an infrared camera, LiDAR, microphones, 3D location sensing and an optical pan, tilt and zoom camera, with on device CPU, GPU and NPU processing; the LiDAR produces three dimensional point clouds mapping patients, furniture and assistive devices; video, depth, audio and location signals are fused on the device before an alert is generated; and the inference target is named as the postural shifts that precede a deliberate attempt to rise, rather than the movement itself. The patient assistant is described as using natural language processing and understanding.

Nothing identifies the model: no model class, no description of training data (a corpus figure the company stated earlier no longer appears on its pages), no version or update practice, and no statement of which inputs the fall and pressure injury models actually use given that the privacy message says no camera is involved.

For the lead function the chain is structurally short: the fall and pressure injury models are the company's own and run on the device in the room, so detection data does not need to go anywhere. The rest of the product is less clear.

Virtual nursing and sitting send video and audio to remote staff, the home page lists virtual nursing with ambient listening and note capture, and the patient assistant uses natural language processing, yet no provider is named for transcription, note generation or language understanding. The privacy policy says personal information may go to service providers such as the company's data centers, without naming them, and no subprocessor list is published.

The company publishes the right kinds of numbers and none of their method, and the numbers do not agree with each other. Lead time is stated as 31 to 65 seconds before a bed or chair exit on the product page and as up to 30 seconds in the Emory Healthcare announcement. Accuracy is 99.99 percent on the home page and over 99 percent for alerts on the product page. False alarm reduction against bed and chair pads is 98 percent on the home page and more than 95 percent on the product page. Outcome figures vary by page too: an 83 percent reduction in falls with injury and over 60 percent fewer hospital acquired pressure injuries in a June 2026 post, more than 61 percent average fall reduction on the return on investment page, and 40 percent fewer falls, 50 percent fewer falls with injury and 40 percent fewer pressure injuries as the calculator's validated deployment averages. None gives a denominator, comparison period or definition.

One hospital case study is summarized on the site: in a Midwest hospital system's neurology unit, over three months with VSTOne in 9 beds, there were no falls with injury and unassisted falls fell by 57 percent (the full paper sits behind a form). Named deployments include Emory Healthcare, where Emory University Hospital Midtown was first to deploy and eight units with about 1,000 beds were planned for 2025. PubMed lists no peer reviewed evaluation of VSTOne; the company's published research concerns VSTBalance, its gait and balance product.

Local processing is a real reduction in exposure: the company says fall and pressure injury detection runs on the device at the bedside, independent of network connectivity, so monitoring data stays behind the hospital's firewall, and the LiDAR sensor produces point clouds rather than photographs. Two things cut against it.

First, the company's privacy policy (updated October 2025), which names VSTOne among the services it covers, reserves the right to use health information and use data for any purpose on a de-identified basis, and to keep de-identified information after an account closes to improve its products. A right to reuse customer data runs against the customer rather than protecting it, and no de-identification method is described.

Second, the privacy message and the hardware do not match. The home and hospital pages say VSTOne monitors without cameras, while the product page lists an infrared camera, microphones, location sensing and an optical pan, tilt and zoom camera, says video, depth, audio and location signals are fused on the device, and offers 4K video streams to virtual sitters and nurses; the product FAQ says live camera feeds are used only for telehealth and virtual nursing. A patient told there is no camera in the room has been told something the product description does not support.

Compliance is claimed and business associate status is not stated. The products page says all VirtuSense products are HIPAA compliant, a July 2026 post calls VSTOne HIPAA compliant by architecture because detection happens on the device, and the privacy policy (updated October 2025), which names VSTOne among the services it covers, says information collected through the services is handled in compliance with the HIPAA Privacy Rule.

That policy reaches the service, but it reads as written for the VSTBalance app's users and care providers, it reserves a right to use health information for any purpose once de-identified, and it says nothing about business associate agreements. No agreement, template or summary of terms is published.

Controls are asserted with nothing independent behind them. The company says monitoring happens locally on the device behind the hospital's firewall, LiDAR data is anonymous by nature, and the system uses end to end encryption, and it sells an off network variant for VA hospitals. No SOC 2 report, ISO 27001 certificate, HITRUST certification, penetration test, trust center or report request path is published. The architecture does narrow the surface for detection, but the same device streams 4K video and audio for virtual nursing and sitting, which a security review has to cover.

No FDA clearance is claimed for VSTOne and none appears under the company's name, and no regulatory position is stated anywhere in its published material. Silence alone is not a failure, and the lead functions, alerting staff that a patient is about to get up and reminding them to reposition a patient who has not moved, have a reviewable basis because a nurse goes to the room and sees the patient. What is marketed goes well beyond that basis.

The company markets autonomous fall and pressure injury prevention with accuracy figures, and its Emory Healthcare announcement says the software also supports virtual nurses by continuously monitoring patient data to detect early signs of deterioration or emergencies, a clinical detection claim with no stated basis. The end user license for its app says the application is not intended to diagnose, treat, cure or prevent any disease, but that document is written for the VSTBalance exercise app.

Nothing is published on how the model's behavior is governed or tested: no subgroup performance, no fairness evaluation, no responsible AI policy, and no description of how the detection model was trained or validated. The company describes its outcome figures and return on investment model as validated, but that refers to deployment results, not to an evaluation of where the model performs well or poorly.

The open questions are specific to movement based inference. Depth sensing does not depend on skin tone the way optical pulse sensing does, but a model reading postural shifts may behave differently for patients with contractures, tremor, amputations, high body mass, mobility aids or restraints in the bed, unusual bedding, or room layouts unlike those it learned from, and for patients whose normal movement in bed looks like an attempt to rise.

A mechanism exists for reviewing what the system did, and nothing stands behind the output. Each alert and safety event is documented automatically in the record, and alerts go to the assigned nurse and through the nurse call system, so a unit can audit the alerts it received and how staff responded. What that record cannot show is the exit the system did not predict, which is the failure that leads to a fall.

Nothing published answers for an error. No sensitivity, missed exit rate or false alert rate is given with a method or denominator: the home page states 99.99 percent accuracy and 98 percent fewer false alarms than bed and chair pads, the product page states over 99 percent accuracy for alerts and more than 95 percent of false alarms eliminated, and neither says what was counted or over how many patients. The published terms of service disclaim warranties, limit the company's liability for direct damages to $50 and ask the user to indemnify the company, and no warranty, indemnity or remediation commitment for hospital customers is published.

Named systems and a described exchange. The company's June 2026 integration post says VSTOne uses HL7 and FHIR data exchange with Epic, Oracle Health (Cerner) and PointClickCare, receives admission, discharge and transfer feeds that switch monitoring on and off per patient, and writes fall risk events, position change alerts and monitoring status into the record without manual entry; alerts can also route through nurse call systems.

At Emory Healthcare it runs virtual nursing integrated with Epic MyChart Bedside, which is evidence of an Epic connection in production. None of this can be checked in a marketplace listing or integration documentation linked from the company's pages, and the product FAQ itself says only that VSTOne integrates with several EHR and other systems.

The deployment is one sensor device per room, no wearable, and no dependence on the hospital's existing monitoring estate. Detection runs on the device's own processors and works without network connectivity, and the company says monitoring data stays behind the hospital's firewall. For VA hospitals it sells VSTSentry, a variant built to run off network.

The virtual nursing and sitting functions necessarily send video and audio to remote staff, and the company's privacy policy refers to service providers such as its data centers, so something runs off site; where, and in which region, is not stated. Alerts route to nurse devices, nurse call systems (Vocera integration announced) and the record.

Named deployments include Emory Healthcare (Emory University Hospital Midtown first, Emory Hillandale Hospital implementing, eight units and about 1,000 beds planned for 2025) and a Midwest hospital system's neurology unit. Residency is stated for the detection function only, and the processing path for virtual nursing video, the assistant and any cloud management layer is left open.

No price is published: no per sensor, per room or per bed cost, no subscription rate, band or implementation fee. What the site does show is how the product is sold and what drives value. It is one sensor per room, so cost scales with rooms covered, and the return on investment calculator states its assumptions (75 percent occupancy, falls with injury at 25 percent of falls, and reduction rates of 40 percent for falls, 50 percent for falls with injury and 40 percent for pressure injuries) and says technology costs are included in its net estimate, which for a 300 bed hospital at national average rates is $1.2 million a year. The cost figure itself is not shown, the reduction rates are the company's own unpublished deployment averages, and the page states the results are illustrative, not a guarantee.

VSTOne is the company's acute care product; VSTAlert and VSTBalance serve skilled nursing, senior living and primary care, and VSTSentry is the VA hospital variant. Within acute care, coverage is any monitored inpatient room rather than a clinical population, and the eight use cases are now listed: fall prevention, pressure injury prevention, active safety monitoring, virtual nursing with ambient listening and note capture, an in room AI assistant (in beta), violent patient alerting, and location tracking with hand hygiene compliance (in development).

Validation sits behind a small part of it. The only hospital result summarized is a three month deployment in 9 neurology beds, and nothing is published on patients or rooms where the sensing works less well, such as patients who move a lot in bed, bariatric beds, restraints or shared rooms.