Inpatient Deterioration & Risk Monitoring

Sibel Health makes soft, flexible sensors that stick to skin and monitor vital signs without a single cable. The ANNE platform pairs a chest sensor capturing electrocardiography, heart rate, respiratory rate, skin temperature, body position and activity with a limb sensor capturing photoplethysmography, oxygen saturation, pulse rate and temperature, feeding an application and central hub that display, alarm and support multi patient monitoring. A Northwestern University spinout founded in 2018, the company sits in the Chicago area with offices in San Diego and Seoul and operates in more than 20 countries. The population coverage is the widest in this index and is backed by separate clearances rather than by claim. ANNE Pediatrics is indicated for neonates including those born extremely premature. ANNE One covers adolescents from 12 and adults, in hospital and at home. ANNE Maternal, cleared in April 2026, is described as the first fully wireless platform for simultaneous maternal and fetal monitoring through labour, delivery and the postpartum period, streaming maternal vital signs alongside fetal heart rate and uterine contraction detection. A separate cough sensor, Aria, has been accepted into a federal qualification programme as a drug development tool, and a Discovery platform serves pharmaceutical clinical trials. Regulatory standing is extensive: at least eight United States clearances, with a seventh announced in March 2025 covering alarms, alerts and a central station, and European Class IIb certification under the current medical device regulation granted in June 2026. Two things distinguish this record from its peers. The sensors are cleared under an open medical device communication standard rather than a proprietary network, and the European certification was announced specifically as the first wireless wearable monitoring platform certified to key interoperability standards. A major patient monitoring manufacturer, which is also an investor and co development partner, cites that open standard as the reason for the collaboration, and the two were selected by the Capital Region of Denmark to deploy across Copenhagen hospitals. The second is bias. The company states it has validated the accuracy of its pulse oximeter across a wide range of skin tones, which addresses the single most documented measurement failure in physiologic monitoring, and that work sits alongside deployments in India, Pakistan, Nigeria and Rwanda with a United Kingdom university research unit, supported by a 17.5 million dollar philanthropic grant aimed at low resource settings. Series C financing reached 39 million dollars by October 2025 with total funding above 63 million, led by existing investors. Two dedicated passes located no pricing of any kind, no information security certification and no trust centre, and the published privacy policy is minimal.

BioIntelliSense replaces the four hourly vital signs check with a coin sized sensor worn on the upper left chest. The BioButton captures heart rate, respiratory rate, skin temperature, body position and activity continuously, generating thousands of measurements per patient per day, which flow through BioHub wireless gateways to BioCloud analytics and surface in the BioDashboard clinical intelligence system. Two device variants exist: a single patient disposable and a rechargeable multi patient version for inpatient use, the latter cleared alongside the dashboard in October 2024. An earlier acquisition brought in a separately cleared clinical intelligence platform. The design decision that defines the product is exception management. Rather than streaming waveforms to a central station, the system holds personalised trending thresholds per patient and notifies only on deviation, which is what allows one clinician to watch hundreds of patients at once. The company publishes the two numbers that make that claim assessable: fewer than one notification per patient per day, and clinician engagement with those notifications at 99.71 percent. Alert fatigue is the recognised failure mode of continuous monitoring, and publishing both the volume and the response rate is the pair of figures that shows alerts are being acted on rather than dismissed. Evidence is genuine and includes an outcome study. Peer reviewed research in the Journal of Clinical Medicine covered nearly 12,000 hospitalised patients across the medical surgical units of two hospitals over 15 months, reporting reduced length of stay, low alert notification rates and earlier intervention on deterioration events. Separately, an independent academic validation published on the device's activity and position detection in children reported sensitivity and specificity figures the company did not commission, including results that are candidly mixed. More than 400,000 patients had been monitored as of October 2025. Deployment includes a full system reference rather than a pilot. Houston Methodist completed expansion across all 2,653 non intensive care beds at its eight hospitals, with a centralised command centre staffed by a multidisciplinary team. Intermountain Health and TriHealth are also named. The commercial argument is built on the alternative rather than on the product, and is unusually well sourced: continuous telemetry costs hospitals up to 1,400 dollars per patient per day, one review found 22.3 percent of patients inappropriately assigned to telemetry at admission with over half monitored longer than guidelines recommend, and a 2024 multicentre study attributed 33 percent of patient sleep interruptions to routine vital sign checks. Based in Denver. Two dedicated passes located no pricing of any kind, no security attestation, no trust centre and no data handling statement.

Vitalchat sells ambient virtual care as software rather than as equipment. The platform performs continuous real time analysis of video, audio and sensory data captured in the care setting, identifying trends, predicting risks and flagging anomalies, which it converts into alerts and, by its own description, automated actions supporting the clinical team. Applications cover artificial intelligence assisted virtual sitting for patient safety monitoring, remote procedural support so a clinician can join an operating room or procedure from elsewhere, automated workflows aimed at nursing administrative burden, and virtual engagement letting family remain present during an inpatient stay. The architecture is the differentiator and the reason this record reads differently from the established vendors in this segment. Vitalchat is a cloud native software as a service platform that leverages a hospital's existing infrastructure and states that it eliminates the need for additional hardware. Competitors in inpatient virtual care sell branded endpoints, install them room by room, and carry device fleet management as a core capability. Removing the capital line item entirely is a genuine commercial and operational distinction, and it is also the claim a buyer should probe hardest, because what the software runs on and what it can see depend on what the hospital already has. Deployment is stated at more than 35 hospitals across medical surgical units, intensive care units, operating rooms and post acute facilities. Commercial launch was in 2021. Funding is 6 million dollars in a Series A closed in February 2025, led by Green Harvest Capital Industries. That investor is a private equity firm whose stated specialisms are multifamily housing, hospitality and industrial assets rather than healthcare, which is an unusual profile for a clinical artificial intelligence company and worth noting rather than glossing, since healthcare specialist investors typically bring diligence and customer access that generalist capital does not. Based in Raleigh, North Carolina and led by chief executive Michael Raymer. A reader should understand what this record does not contain. Two dedicated passes located no security page, no external attestation, no health privacy statement, no pricing of any kind, no named customer, no named record system integration and no published performance figure for any model. The company describes what its technology does at the level of capability and publishes almost nothing that would let a buyer verify or evaluate it. That gap, rather than the technology, is the substance of most grades in this record.

Caregility runs the video layer of the hospital and is adding intelligence on top of it. Caregility Cloud, now presented as the Caregility Connected Care Platform, delivers secure two way audio and video into patient rooms across inpatient, ambulatory, post acute and home settings, and applications sit on that layer for specific jobs: iObserver for virtual observation and e sitting, iConsult for virtual consultation, and iCare Coordinator for coordinating bedside clinicians, remote caregivers and AI tools in one place. The scale is larger than any comparable vendor in this index. More than 1,500 hospitals and 75 health systems across multiple continents, over 30,000 connected devices, and more than six million virtual sessions a year. Programmes span virtual nursing, virtual observation, virtual rounding, specialty consults, operating room telehealth and hospital at home. The artificial intelligence is real and is being built rather than inherited. Edge based computer vision runs on the devices, audio sensing detects room duress, and ambient listening and sensor based capabilities are in the platform, with an AI enhanced unlimited patient monitoring capability added to the coordination application in 2026. The 25.1 million dollars raised in September 2025, led by Star Mountain Capital and bringing total outside investment to 92 million, was explicitly earmarked for computer vision, ambient listening and sensor technologies. A Chief Product, Technology and AI Officer holds the portfolio. Integration is the strongest part of the record. Caregility has been listed in Epic Toolbox for Inpatient Virtual Care for two consecutive years including 2026, with named integration paths into MyChart Bedside TV, Epic Monitor, Haiku and Hyperspace, alongside nurse call systems, interactive patient consoles and bedside devices. Ecosystem partners run their own applications on the platform, with a clinical documentation vendor among those named. Connectivity options extend to 5G and satellite for rural and mobile deployment. Compliance disclosure is also stronger than the segment norm. The platform is stated to be HIPAA, GDPR and CCPA compliant and ISO and SOC certified, and a named independent security firm is cited as having verified the HIPAA position. The platform has been recognised as the leading virtual care platform outside the record system vendors by a third party analyst firm in three consecutive years. Based in Wall, New Jersey. One thing a buyer should weigh directly. The footprint here is larger than the closest competitor's and the published evidence is thinner: no peer reviewed study was located, and named customer results are testimonial rather than quantified, where the closest competitor publishes response times, alarm reductions and dollar figures at named institutions. Scale and evidence are different things and this record separates them deliberately.

AvaSure invented virtual sitting and has spent fifteen years turning it into an enterprise platform. A camera in the patient room feeds a remote observer who can speak to the patient through two way audio and summon staff, and computer vision watches alongside that observer, flagging potential falls and other high risk events so attention lands where it is needed. The ratio is the product: a single observer safely covers up to 16 patients, which is what converts one to one sitting into something a hospital can staff. The platform now spans continuous observation, virtual nursing, episodic consults, virtual visits, ambient sensing and an AI agent for virtual care workflows. Hardware comes in fixed and mobile forms, with 1080p cameras carrying 10x optical zoom and 360 degree pan, tilt and zoom, infrared illumination for night vision, and a dual camera variant pairing a wide angle lens for ambient signal capture with a second lens for video consults. Ceiling mounted units are tamper resistant and available in ligature safe configurations for behavioral health. Integration is the part that separates this from a camera system. The virtual visit application launches from within Epic Hyperspace, devices pair with Epic Monitor for virtual observation, and urgent alarms route to nurse call systems in real time over HL7 interfaces, application programming interfaces or middleware. Caregiver acknowledgements then sync back into AvaSure, which closes the loop and makes response measurable rather than assumed. Scale and evidence are both unusual for this index. More than 1,200 hospitals are deployed. A peer reviewed study published in early 2026 evaluated a continuous virtual monitoring programme in a long term acute care hospital across 40 months, reporting sustained safety and financial outcomes. Named customer results include a 38 percent reduction in urgent alarm rates with an 11 second average response time at one health system, and a 16 percent cut in emergency department holding time with a 31 percent improvement in discharge turnaround at an urban safety net hospital. More than 150 customer generated presentations and studies exist, and one health system published 3.2 million dollars in measurable impact. The company began in Grand Rapids, Michigan as a business security integrator and launched the TeleSitter line in 2009, adding a team of registered nurses who run the clinical education programme that accompanies the technology. Importantly for membership here, AvaSure sells the platform and the hospital staffs the observers, which is what distinguishes it from managed monitoring services that supply observers from their own centre. One disclosure worth knowing before procurement. AvaSure publishes no rate card, but a United States Department of Veterans Affairs solicitation sets out the commercial structure in the open: software licensed per hardware device on an annual term at a named tier, with support tiered separately, and hardware, clinical services, monitor stations and installation carried as distinct line items. A federal variant of the hardware carries validated cryptography.

Ambient artificial intelligence and computer vision for the operating room. Sensors installed in the room observe surgical activity, machine learning models automatically identify up to fourteen distinct case events, and the resulting structured data is written autonomously back into the electronic record and fed into scheduling, utilisation and turnover analytics. Founded 2020 by David Schummers, chief executive, and Cameron Marlow, chief technology officer. Headquartered in San Francisco, with some sources giving Oakland. The rationale is concentration of value rather than breadth. The company states that up to 60 percent of a health system's revenue and 40 percent of its costs sit in the operating room, which makes perioperative capacity the highest leverage target in hospital operations. A later product, Horizon, forecasts case duration and utilisation to optimise scheduling before the day begins rather than reacting during it. Funding totals 101 million dollars: a 21 million dollar Series A in December 2021 led by Casdin Capital, and an 80 million dollar Series B of equity and venture debt announced 8 January 2026 led by HighlandX, with returning investors Vensana Capital, Casdin Capital, PFM Health Sciences, Upside Partnership and Operator Partners, and new investors K2 HealthVentures, OpAmp Capital and Houston Methodist. The president of the Johns Hopkins Health System joined the board. The flagship deployment is Houston Methodist, a nine hospital system that ran an initial 36 room pilot and has since scaled to more than 200 operating rooms enterprise wide, with its executive vice president and chief innovation officer quoted on the result. That reference carries a disclosed conflict this record records plainly: Houston Methodist is also an investor in the Series B, and the quoted executive appears among the company's individual investors. The evidence is real and the commercial relationship runs both ways. The headline outcome claim is an average 5 percent increase in surgical volume across customers, vendor stated with no methodology or study attached. Two characteristics distinguish this record from everything else in the lane and drive several grades. The system writes machine generated observations autonomously into the legal medical record, which is a different act from advising a human. And the data it collects is continuous video of surgical procedures, capturing patients under anaesthesia and the staff around them. Neither is addressed in any published material located, and both are recorded on the relevant axes. Disclosure limitation: a dedicated pass located no security certification, trust page, privacy posture, integration specification or pricing information. The trust axes are graded conservatively on that basis and should be revisited.

Virtual care collaboration platform repositioned around agentic artificial intelligence, sold as ThinkAndor. Founded 2018 in Orlando, Florida. Microsoft's venture arm took an investment position in 2020 and the two work closely on the underlying model technology. The platform is organised around five pillars of virtual health: virtual visits, virtual hospital, virtual patient monitoring and care management, virtual team collaboration, and virtual community collaboration. In practice that spans device agnostic virtual rounding, virtual nursing and virtual sitting, remote specialty consults including tele stroke, tele psychiatry and tele intensive care, smart room deployments, digital front door agents and conversational patient interaction. The company describes generative capability for ambient sensing and conversational documentation, pulling patient and clinical context out of source systems into the virtual encounter and pushing recommended tasks back into the record. Third party validation is the strongest part of this record and is unusually well quantified. ThinkAndor was rated 2026 Best in KLAS for virtual care platforms in the non record system category, with a published score of 95.9 out of 100, which the company states is 15.7 points above the overall average and 15.1 points above the virtual care platform average. Black Book has rated it the highest scoring virtual care platform in the same category for three consecutive years, and Frost and Sullivan recognised it in 2025 for acute care virtual health. Two national group purchasing agreements are in place, with Vizient from February 2025 and Premier from February 2026. Scale is stated at more than 70,000 providers and more than 500 hospitals across the United States, Canada and the United Kingdom. Named deployments include Sentara Health across twelve hospitals from November 2025, Providence, and the National Institutes of Health Clinical Center. Published outcome claims are vendor stated and are not backed by any study located in this pass: ten to twelve minutes saved per visit, a 35 percent reduction in call abandonment, reduced call centre utilisation, and a 47 percent reduction in readmissions attributed to intelligent orchestration of patient care. The readmission figure in particular is a substantial clinical claim carrying no published evidence, and should be treated as marketing until a source is produced. One sourcing note. The clearest statement of the underlying model technology, that the Microsoft relationship optimises virtual interactions using that company's partnered large language models, appears on a third party health system marketplace listing rather than in the company's own material, and is recorded on that basis.

care.ai turns patient rooms into sensed environments for virtual nursing and virtual monitoring, using camera based sensors with graphics processing units on the unit itself, so computer vision runs at the bedside rather than in a distant cloud. From a single virtual command centre, remote nurses or sitters support the bedside team, and Stryker states that one remote nurse can monitor, assess and support up to ten patients at once using the real time data the ambient system produces. The company was based in Orlando, Florida and was acquired by Stryker in 2024 for an undisclosed sum. It no longer trades independently, and this record is scoped to the care.ai product line, which Stryker continues to market under that name. What has changed since is the context around it. In March 2026 Stryker launched its SmartHospital Platform, run by a newly formed Smart Care business, which brings care.ai's ambient intelligence and virtual care workflows together with more than 85 Stryker products on one interface. Those include the ProCuity bed, the Vocera hands free communication devices and Sync Badge that Stryker acquired for close to 3 billion dollars in 2022, and the Engage middleware engine that filters and prioritises alarms. Vocera alone integrates with more than 150 clinical and operational systems including record systems, nurse call, ventilators and physiological monitors. So a hospital evaluating this is no longer evaluating an ambient intelligence vendor. It is evaluating one module of a medical device manufacturer's connected ecosystem, sold modularly but designed to be adopted together, and the more of that ecosystem a hospital already owns the more the ambient layer is worth to it. Anyone searching for care.ai or Vocera is looking at Stryker.

Mednition sells KATE, a clinical artificial intelligence platform that works at the emergency department triage desk rather than on the ward. It was founded in 2014, is based in Burlingame, California, and is led by chief executive Steven Reilly. KATE reads the structured intake data a triage nurse enters together with the free text of their notes, compares it against patterns from millions of prior visits, and recommends an acuity level in real time. The company is explicit that it supports rather than replaces the nurse's judgement, adds no new screens and requires no workflow change, which matters because triage is the most time pressured decision point in the hospital. The sepsis model is the flagship. KATE Sepsis received Breakthrough Device Designation from the Food and Drug Administration in November 2023 for detecting sepsis at triage, before any laboratory result exists. It has not been cleared: the designation expedites review rather than authorising marketing, and no clearance was located nearly three years later. Performance is published in preprints with named authors rather than only in marketing copy. An early model reported an area under the curve of 94 percent for sepsis detection at triage. A later model reported an area under the curve of 99 percent with 95 percent sensitivity and 96 percent specificity, on a retrospective cohort of 540,884 patients containing 14,676 positive sepsis cases across 16 hospital sites, using the current academic sepsis definition. Against standard screening protocols the company reports sensitivity improvements of 74 percent for sepsis, 80 percent for severe sepsis and 118 percent for septic shock. The company names the failure mode its category is known for rather than avoiding it, stating that the challenge has been achieving high sensitivity without a collapse in specificity and the false positive rates and alert fatigue that follow. Adventist Health has adopted KATE systemwide, and named sites include Shady Grove, White Oak, Fort Washington and Glendale. A separate paediatric safety product covers the same triage moment. KATE was named best in show in the capacity crisis category at the 2025 HIMSS conference.

Prenosis is a Chicago company whose Sepsis ImmunoScore was the first artificial intelligence diagnostic for sepsis ever granted marketing authorisation by the Food and Drug Administration, cleared through the De Novo pathway on 3 April 2024. Because De Novo creates a new device classification, that authorisation established the category a later competitor used as its predicate, which makes this record the regulatory origin point of the whole cleared sepsis segment. What separates it technically from the other deterioration products in this index is that it does not read the record alone. The score combines biological markers measured from a blood sample with clinical data drawn from the record, using up to 22 parameters, and returns a risk score placing the patient in one of four discrete risk categories. Those categories are tied to length of stay, in hospital mortality and escalation of care within 24 hours, meaning intensive care admission, mechanical ventilation or vasopressor use. The company states explicitly that it is not an alert system. The underlying asset is the Immunix platform and the biobank built on it: more than 100,000 blood samples from over 25,000 patients, assembled across a decade with ten partner hospitals and held in a biosafety level 2 laboratory in Chicago, paired with clinical data from those hospitals' records. The company describes this as the largest combined biological and clinical dataset in the world for acute care patients suspected of serious infection. Sepsis ImmunoScore is distributed commercially through a collaboration with Roche, appearing on the navify Algorithm Suite. In January 2026 the company announced 40 million dollars, comprising a 20 million dollar Series A led by PACE Healthcare Capital with the Labcorp Venture Fund and Carle Health, and a 20 million dollar federal contract from BARDA funding a randomised controlled trial of 800 patients with severe respiratory infections. Co founder and chief executive Bobby Reddy Jr. Founding year was not confirmed in this pass and is left blank.

Artisight sells a smart hospital platform that turns a patient room into a sensed environment. A multi sensor network of cameras, microphones and positioning hardware runs computer vision and speech models on graphics processing units installed on site, and the system continuously observes the room and responds to what it sees and hears. The capability list spans several jobs the index usually treats separately: virtual nursing, virtual observation in place of an in person sitter, fall prevention, ambient documentation of room activity, two way audio and video for tele consultation, vital sign monitoring, indoor positioning, and surgical case length prediction in the operating room. The company describes itself as an ambient intelligence platform deployed across clinics, patient rooms and operating rooms rather than as a point solution. It was founded in 2015 by Andrew Gostine, a physician, and is based in Chicago. It raised a 42 million dollar Series B in January 2024 that was oversubscribed and included NVIDIA among the investors, followed by a further 40 million dollars reported in August 2025, at which point the platform was described as in use at 417 hospitals including Northwestern Medicine, WellSpan and the Guthrie Clinic. A notable feature of the capital structure is that many client hospitals are also investors. Northwestern Medicine is cited as reporting an 89 percent reduction in falls, a 52 percent reduction in nursing overtime and a 76 percent reduction in nursing turnover. In 2024 the chief executive stated that publication of these results in the peer reviewed literature was anticipated. No such publication was located in this pass.

Clinical deterioration early warning and clinical pathways company built on research from the University of Chicago, indexed primarily on eCART, its FDA cleared early warning system. eCARTv5 is a cloud based gradient boosted machine learning model, integrated into the EHR, that synthesises routine vital signs, laboratory data and patient demographics into a single score predicting the composite outcome of death or ICU transfer for adult ward patients. The 510(k) summary discloses the model class outright, publishes the default alerting thresholds of 93 for moderate risk and 97 for high risk on a 0 to 100 scale constructed from specificity, and states that the observed rate of deterioration at each threshold is displayed to the clinician as odds of deterioration in the next 24 hours alongside the score, which puts calibration in front of the user at the point of care. The cleared indication restricts the model to data a healthcare professional has already validated, so unconfirmed streaming data from monitors and devices is excluded until a nurse or clinician confirms it. Clearance rested on retrospective validation across 1,769,461 encounters and prospective validation across a further 205,946 encounters in three geographically distinct health systems, and the filing publishes performance stratified by five race categories and by comorbidity, which no other vendor in this category publishes anywhere. The company separately sells Clinical Pathways, a content and workflow product with more than 4,800 pathways live across adult, paediatric and neonatal topics in ambulatory, emergency, inpatient, ICU and women's health settings, which serves as the response layer the cleared eCART workflow directs clinicians into. Co founded by Dr Dana Edelson, Chief Medical Officer, whose University of Chicago group produced the underlying research across more than fifty peer reviewed publications since 2011. CEO and co founder Borna Safabakhsh.

HealthLeap runs continuous AI screening across every hospitalised patient, starting with malnutrition. The problem it targets is a screening failure rather than a detection failure: hospitals typically assess nutritional risk through a questionnaire completed by nursing staff on admission, which is static, so a patient who is adequately nourished on day one and declining by day four is missed. Nationally, the company cites 30 to 50 percent of hospitalised patients as being at risk of inadequate nutrition while fewer than 9 percent are formally diagnosed. The platform runs in the background of the electronic health record from admission through discharge, analysing the full chart daily rather than at a single point: clinical notes, laboratory results, vital signs, active medication orders and problem lists. Risk scores and prioritisation alerts are pushed into the native EHR interfaces that nurses and dietitians already use rather than into a separate application. The company describes malnutrition as the first condition rather than the only one. HealthLeap publishes a dedicated clinical use and safety statement that is more explicit than most in this category. It states that the software is not a diagnostic tool, does not provide treatment recommendations, must not be used as the sole basis for determining nutritional risk or the presence of malnutrition, and should be used alongside clinical assessment, physical assessment and laboratory findings. It also states a clear technical boundary: the software uses only existing EHR data and evidence based risk factors, and does not acquire or process medical images, waveforms or signals from medical devices. The commercial case is made in two halves and the second is worth reading carefully. Alongside earlier identification and shorter length of stay, the company positions the platform as improving documentation accuracy so that the record reflects a patient's true severity of illness, aligning frontline care with hospital coding, optimising reimbursement accuracy and defending against retrospective payer claim denials. HealthLeap is based in San Francisco and announced a systemwide deployment across Houston Methodist covering more than 150,000 inpatients annually. It claims to be the only commercially available peer reviewed validated platform of its kind, a claim this record was unable to verify against a retrieved publication. Pricing is not published.

Kinometrix builds machine learning models that predict hospital acquired harm from data already in the electronic health record. Its first product, the K-FRAS fall risk assessment system, is a proprietary predictive model producing real time fall risk predictions with the specific risk drivers attached, delivered to the frontline clinician. A pressure injury product follows the same pattern. The company's argument for why a model is needed is sharper than most in this category and is worth stating in full, because it describes a failure mode rather than an opportunity. Nurses currently complete manual fall risk scores. Those instruments lack specificity and systematically overestimate risk, because they are built to avoid missing an at risk patient. The consequence is that a large proportion of patients get labelled high risk, and once that happens an organisation cannot resource individualised prevention, so it falls back on generic precautions applied to everyone. Kinometrix's position is that a more specific model lets a hospital direct resource intensive interventions to the patients who actually need them rather than diluting them across a ward. Architecturally the platform is described as headless, meaning it carries no interface of its own and can be integrated with any EHR and configured to a given hospital's needs. It runs in the background against the record, evaluates markers to build a risk profile, writes updates back into the EHR automatically, and surfaces recommended interventions matched to that profile. The company emphasises that it adds no documentation burden, in contrast to tools that require nurses to complete additional fields, and the model takes the nurse's own expert assessment as one input alongside objective record data rather than replacing it. Published claims include accuracy of around 98 percent, elimination of false low risk assessments, and a 6.5 to 1 return on investment, none of which is accompanied by a derivation, denominator or independent evaluation. Context the company cites for the problem: roughly one million hospitalised patients fall each year in the United States, about a third of those falls cause injury, and the annual cost to US hospitals is given variously as 6 and 7 billion dollars across its own pages. Pricing is not published.

VSTOne is VirtuSense's acute care product, an in room sensor system that predicts when a hospital patient is about to leave their bed or chair unassisted and alerts staff before it happens. It belongs in this category for a reason worth stating plainly: bed alarms are the canonical alarm fatigue problem in hospital nursing, and this product is built and sold entirely in those terms. A single LiDAR based sensor is deployed per room, described as part of a multi modal stack that also includes infrared imaging, microphones, real time location sensing and an optical pan tilt zoom element. Computation runs locally at the bedside rather than transmitting video to external servers. The system interprets spatial data as clinical context rather than discrete motion events, identifying the subtle postural shifts that precede a deliberate attempt to stand, and issues an alert roughly 31 to 65 seconds before the exit occurs. That advance warning is the entire product thesis: a conventional bed pad alarm fires once the patient has already gone, which is too late to prevent anything. VirtuSense positions VSTOne against the performance of existing bed and chair pads directly, stating that conventional pad accuracy can be as low as 15 percent, that nursing staff encounter 20 to 30 false alarms per bed each day, and that its own system achieves 98 percent fewer false alarms and a 99.99 percent accuracy rate, with organisations seeing an average 74 percent reduction in falls with injury. None of these figures is accompanied by a denominator, a definition or a published method. The product is described as covering eight use cases from one device, with fall risk and pressure injury named explicitly, and as fully integrated with Epic, with alerts logged in the platform and written to the electronic health record. Group level reporting covers fall trends by unit or site over time. The company's earlier products remain separate: VSTBalance, launched in 2014, performs a two minute LiDAR based gait, balance and function assessment, and VSTAlert serves skilled nursing and assisted and independent living. This record is scoped to VSTOne, the acute inpatient product, consistent with this index's practice of indexing by product and with the setting boundary that places residential and long term care fall monitoring outside this category. Pricing is not published.

Biobeat sells wearable continuous vital sign monitors and a cloud platform, and its position in this category is distinctive: where other vendors build models on data that already exists, Biobeat changes what data exists in the first place. Its answer to the ward monitoring problem is to make ward patients continuously monitored rather than measured every few hours. The hardware is a disposable chest patch and a wrist worn device, both built on a proprietary reflective photoplethysmography sensor. From that single optical signal the company derives a set of parameters that has expanded through successive FDA clearances: cuffless blood pressure, blood oxygen saturation and pulse rate cleared in August 2019, respiratory rate and body temperature added in March 2022, and stroke volume and cardiac output added in October 2024. Biobeat states these were the first devices cleared by the FDA for cuffless blood pressure monitoring derived from photoplethysmography alone. The devices also carry CE marking and the company holds MDSAP certification. Data transmits to a cloud platform for viewing by clinical staff, with integration into the hospital EMR. On top of the measurements sits an automated real time early warning score which the company describes as incorporating advanced AI based algorithms to alert on patient status and potential deterioration. The FDA clearances cover the measurement of physiological parameters; nothing retrieved indicates the early warning score itself is cleared, and a buyer should establish that distinction directly. Marketed settings span general wards, surgical wards, post surgical recovery, step down and medical surgical units, intensive care and emergency departments, alongside ambulatory blood pressure monitoring, nursing homes and hospital at home programmes. The chest patch is also used for 24 hour ambulatory monitoring with automated report generation. Biobeat was founded in 2016, is headquartered in Petah Tikva, Israel, and is led by founder and chief executive Arik Ben Ishay. International distribution agreements include partners in the Dominican Republic, Argentina and Chile. Pricing is not published.

Sickbay, from Medical Informatics Corp, is not a risk model. It is the data layer that risk models run on, and it is included in this category because it is FDA cleared for inpatient patient monitoring and alarm analytics and because its argument bears directly on how every other vendor here should be evaluated. The company's position is that the substrate is the problem. Predictive analytics in hospitals are typically built from discrete data points in the electronic medical record, which MIC argues are incomplete, significantly delayed and subject to human error, so predictions arrive after deterioration or are simply wrong. Sickbay's answer is to capture high fidelity, time series waveform data directly from bedside devices in near real time, at a stated average resolution of around 25 milliseconds per patient, and to persist it. Medical device manufacturers store data in proprietary formats, so the platform's central claim is vendor neutrality: it unlocks waveform and vitals data across disparate manufacturers, including from non networked devices such as ventilators, and makes it available through a browser on laptops, mobile devices, wallboards and in on premise or remote command centres. Crucially for how it is graded here, Sickbay supplies data to clinicians, researchers and algorithm developers through APIs and development tools rather than supplying its own deterioration model. Risk indicators in the Virtual Ops module are configurable by the institution. The platform received FDA 510(k) clearance as a Class II device along with three applications: Patient Monitor, Patient Alarm Data and Alarm Analytics Dashboard. The last of those is the only cleared product in this category aimed specifically at measuring alarm burden, which is this category's defining failure mode. Current modules are Sickbay core, Sickbay Telemetry, Sickbay Virtual Ops and Sickbay Analytics, with a dedicated rural healthcare offering. The company states any bed can be scaled to a monitored ICU bed within minutes through a database change. MIC is based in Houston and led by chief executive Emma Fauss, with Craig Rusin as chief product and innovation officer, whose research produced the grid computing platform underlying the product. Funding includes a Series A led by DCVC with Intel Capital and the Texas Medical Center Venture Fund, and a later 27 million dollar round co led by Catalio Capital Management and Intel Capital. Named institutional relationships include Texas Children's Hospital and Tampa General Hospital. Pricing is not published.

AlertWatch:OB is a maternal safety surveillance system for labor and delivery, and its distinguishing feature is that alarm fatigue was the founding design problem rather than an afterthought. The company was founded in 2012 as a University of Michigan spinout by Dr Kevin Tremper, chair of the university's anesthesiology department, and the founding team included Dr James Bagian, a former NASA astronaut who was the founding director of the VA National Center for Patient Safety and the first chief patient safety officer for the Veterans Health Administration. Tremper has described designing the system around aviation cockpit principles and the lesson of Three Mile Island, where every alarm sounding at once proved distracting rather than useful. The product monitors mothers rather than fetuses, and covers an unusually wide span of the obstetric journey: from triage through labor and delivery, into operating rooms and post anesthesia care units, and through the postpartum period. It integrates hundreds of data elements from physiological monitors, the EMR, laboratory systems and medical history, and alerts clinicians using a proprietary maternal early warning score based on national obstetric standards, with proprietary filtering intended to raise alerts without producing alarm fatigue. It assesses ACOG postpartum hemorrhage risk every minute for every mother, and, notably, checks automatically whether blood is available and whether intravenous access is adequate, so the system tracks the hospital's readiness to respond rather than patient risk alone. AlertWatch:OB is FDA 510(k) cleared and the company describes it as the first and only cleared software system dedicated solely to maternal safety in labor and delivery. It follows AlertWatch:OR, cleared in 2014 for the operating room, and AlertWatch:AC for acute care. Commercial launch of the OB product was announced in January 2020, at which point it had assisted with more than 10,000 births, rising to more than 15,000 in reporting later that year. Distribution has run partly through Clinical Computer Systems, maker of the OBIX electronic fetal monitoring system, as both reseller and co marketing partner. The company retains a relationship with the University of Michigan covering intellectual property and ongoing product testing. AlertWatch was acquired by BioIntelliSense in 2022. Pricing is not published.

PeriGen sells the PeriWatch platform, an early warning and clinical decision support system for labor and delivery. It is the obstetric member of this category, and the only vendor assessed here whose algorithms were validated by experts at a federal research institute rather than by the company or its customers. PeriWatch Vigilance continuously monitors both the mother and the fetus, drawing from the hospital's existing electronic fetal monitoring system and EMR, and notifies clinicians when maternal vital signs, fetal heart rate patterns or labor progress assessments cross hospital defined safety limits. It escalates to designated hospital leaders when thresholds are breached, and is accessible through an obstetric command center view and on mobile. The architecture is a hybrid worth noting: the fetal heart rate pattern interpretation and uterine contraction detection use supervised machine learning, which the company states plainly, while the alerting thresholds themselves are set by the institution rather than the vendor. The underlying algorithms have a long regulatory and research history. Patterns performs automated fetal heart rate pattern recognition and Curve assesses labor progression, accounting for factors including contraction frequency and epidural use. The company describes these as the only FDA cleared algorithms for fetal heart rate interpretation and labor progress assessment commercially available in the United States. Patterns 3.0 was cleared in February 2025, extending the indicated range from 36 weeks of gestational age down to 32 weeks to cover earlier monitoring of high risk pregnancies. PeriWatch Surveillance is separately 510(k) cleared. Independent validation came from three experts at the Eunice Kennedy Shriver National Institute of Child Health and Human Development, who reviewed the software's analysis across 100 tracings and reported agreement in over 97 percent of assessments, concluding that computerised fetal heart rate interpretation shows substantial agreement with expert evaluation and can screen in real time when an expert is not continuously watching. Separate peer reviewed work in the American Journal of Obstetrics and Gynecology compared the software against five expert clinicians using a strict classification framework and found the computer results not statistically different from the clinicians. PeriGen is based in Cary, North Carolina, is led by chief executive Matthew Sappern, and is a Halma company. It acquired the WatchChild fetal monitoring system from Hill-Rom. Named deployments include Mount Sinai Medical Center, Avera Health and NYC Health + Hospitals. Pricing is not published.

Healthplus.ai sells PERISCOPE, a CE certified clinical decision support system that predicts the risk of postoperative bacterial infection. It carries the most methodologically complete published evaluation of any vendor in this category, and its central research contribution is the problem this category keeps running into: that a model developed at one hospital does not necessarily work at the next one. PERISCOPE reuses data already in the electronic health record, requiring no new measurements, and draws on roughly 50 clinical parameters covering the preoperative and intraoperative period. It produces two predictions per patient, one for infection within seven days of surgery and one within 30 days, presented to surgical teams as three traffic light categories inside the EHR workflow. The scope covers postoperative bacterial infections broadly rather than surgical site infection alone, including pneumonia, urinary tract infection and other bacterial infections. The models are XGBoost, and the company states they are calibrated to each hospital's own data. The validation work was published in The Lancet Regional Health Europe in December 2024. Models were developed at one hospital and then validated and updated at two further hospitals in the Netherlands and Belgium, across 253,010 surgical procedures using data from 2014 to 2023 spanning multiple surgical specialties, with the two most recent years held out for temporal validation. Performance was reported on discrimination, on calibration including slope, intercept and plots, and on clinical utility through decision curve analysis with net benefit. The study was funded by a REACT EU grant from the European Regional Development Fund rather than by the company, and conflicts are disclosed: the chief executive is a major shareholder and two authors are employees, while the substantial academic co author list declares none. The product holds ISO 13485 and CE certification as Class IIa software as a medical device under the EU Medical Device Regulation, a route that requires clinical evaluation. The company has stated it is licensed to operate in the EU with initial focus on Benelux and Germany while pursuing FDA authorisation for the US. Named clinical sites include Amsterdam UMC and Deventer Ziekenhuis. A 2.3 million euro round was led by Elevating Capital and LUMO Labs with Pathena Venture Capital, Leistone and ROM InWest. Pricing is not published.

Luminare sells inpatient sepsis screening and intervention software, and it is the only vendor in this category whose founding argument is that better prediction is not the problem. The company's position, stated on its own site and argued in a peer reviewed review article co authored by its chief executive, is that the health technology industry has concentrated on speed of detection while outcomes stall because clinical staff frequently fail to act on the alert. Its answer is not a stronger model but an automated intervention workflow. The mechanism is correspondingly transparent. Luminare screens patients on admission through EMR integration and then once per shift, combining information already in the medical record with the nurse's own assessment, structured around what the company calls an enhanced SIRS screening checklist. When criteria are met it drives the hospital's existing sepsis protocol, issuing clear intervention steps in SBAR format, routing inter departmental communication, and reporting performance against the hospital's own benchmarks. It also supports SEP-1 bundle documentation for the three and six hour bundles. Notably, the nurse's assessment is an input to the screen rather than a review step applied afterwards, so unlike a background risk score the output cannot be passively ignored. The company was founded in 2014 in Houston by Sarma Velamuri MD, a board certified internal medicine physician and hospitalist, and Marcus Rydberg, following the death of a friend's daughter from septic shock. It is based at the Texas Medical Center Innovation Factory. Luminare signed an enterprise agreement with Cedars-Sinai in 2023 after participating in that health system's accelerator, and names Microsoft, Cerner, CPSI and the American Heart Association Get With The Guidelines programme as interoperability partners. The company reports 36 staff. Pricing is published as a three tier structure with the unit of pricing named: a standalone product requiring no integration offered on a 30 day trial, an Enterprise Lite tier on a flat annual fee deployable in weeks, and an enterprise platform priced per patient day on an annual contract and implemented over months. A separately published grant programme offers three, six or twelve month deployments at no financial cost to qualifying smaller facilities, in exchange for data sharing, focus group participation and testimonials. Luminare is not FDA cleared.

AITRICS is a Korean medical AI company whose AITRICS-VC (VitalCare) predicts patient deterioration from electronic medical record data. It is the first vendor in this category focused on the general ward rather than the ICU, which is a materially different detection problem: ward patients are observed intermittently rather than monitored continuously, so the model must work from sparse, irregularly timed measurements. As marketed, VitalCare runs two deep learning models built on a bidirectional long short term memory architecture. VC-MAES predicts clinical deterioration events, defined as unplanned ICU transfer, cardiac arrest or in hospital death, within six hours. VC-SEPS predicts sepsis onset within four hours. The company also markets cardiac arrest prediction within 24 hours and, in the ICU, mortality prediction within six hours. Inputs are 19 parameters drawn from the EMR: six vital signs, 11 blood test results, level of consciousness and age. The development and validation work is published as a medRxiv preprint. Models were derived on 357,009 adult general ward admissions at Yonsei Severance Hospital between 2013 and 2017, then externally validated on 22,073 admissions at National Health Insurance Service Ilsan Hospital. In external validation VC-MAES reached an AUROC of 0.918 against 0.834 for MEWS and 0.883 for NEWS, and VC-SEPS reached 0.941 against 0.559 for SOFA, 0.687 for qSOFA and 0.767 for NEWS. Both models held AUROC above 0.86 across all age and sex categories, which is the only published subgroup performance reporting located for any vendor in this category. Regulatory footprint spans several jurisdictions, and the scopes differ in ways a buyer must read carefully. Korea's MFDS approved the deterioration prediction product for both general wards and intensive care. Hong Kong's Medical Device Division and Vietnam's Ministry of Health followed. The US clearance is materially narrower: 510(k) K240756, granted 23 July 2024 under 21 CFR 870.2300, describes software performing rule based calculation of conventional early warning scores including NEWS, MEWS and qSOFA, screening patients against predefined thresholds and displaying them on a dashboard. It is indicated solely for the general ward and is expressly not indicated for the ICU or operating rooms, and the submission required no clinical data. VitalCare is reported in use at more than 60 hospitals in Korea. Pricing is not published.

Etiometry sells a clinical intelligence platform for critical care that combines vendor neutral ICU data aggregation with four FDA cleared physiologic risk indices. It is the most technically disclosed vendor assessed in this category, and it takes a fundamentally different approach from the machine learning classifiers that dominate the lane. The indices do not predict a coded outcome. They estimate the probability that a patient is currently in a defined physiologic state, and each state is specified against a published clinical threshold. IDO2 estimates the likelihood of inadequate oxygen delivery against a configurable mixed venous oxygen saturation threshold. IVCO2 estimates the likelihood of hypercapnic respiratory failure. HLA estimates the likelihood of hyperlactatemia, defined as lactate above 4 mmol per litre. ACD estimates acidemia, defined as arterial pH below 7.25. Each runs continuously using what the company describes as a Bayesian modelling approach combining mechanistic models of human physiology with techniques drawn from aerospace navigation, taking heart rate, blood pressure, SpO2 and, where available, filling pressures, venous oxygen saturation, hemoglobin and blood gases. The models are built to tolerate missing or intermittent inputs, which the company identifies as a common failure point for simpler rule based early warning scores. Around the indices sits a data layer with unusual reach. Etiometry ingests numeric data at five second intervals and waveforms up to 500 Hz from bedside monitors across GE Healthcare, Philips, Draeger, Mindray, Nihon Kohden and Spacelabs, ventilators from Medtronic, Draeger, Hamilton Medical and Getinge, ECMO, VAD and CRRT systems, cerebral oximetry and hemodynamic monitors, using medical device integration platforms including Capsule, Bedcomm and IBUS. EHR data arrives from Epic, Oracle Health and Meditech over HL7 and a FHIR application with SMART on FHIR single sign on, and derived risk indices can be pushed back out over HL7 and REST. A Quality Improvement Application retains ingested data with full fidelity waveforms archived separately, and supports pathway adherence reporting and event reconstruction. The company reports 11 FDA clearances covering the platform, persistent bedside display functionality, the four risk indices and, in 2026, the Cardiogenic Shock Tool under K254066, alongside EU MDR, Health Canada and MDSAP authorisations, ISO 13485, ISO 27001 and SOC 2 Type II. Deployment is stated at roughly 150 ICUs and 3,500 beds across 50 hospital partners. The published evidence base is concentrated in paediatric cardiac critical care and includes a multicentre before and after study in Critical Care Medicine and work in Circulation: Cardiovascular Quality and Outcomes. Pricing is not published.

CLEW Medical sells FDA cleared machine learning models that predict clinical deterioration in hospital critical care, delivered through a virtual ICU platform. It is the only vendor assessed in this category whose absolute operating characteristics are published in a public document, because the models are regulated as medical devices and the 510(k) summaries state them. The cleared device, the CLEWICU System, contains two models. CLEWHI estimates the likelihood that a patient will develop hemodynamic instability requiring vasopressor or inotrope support within the next eight hours. CLEWLR indicates that a patient is at low risk of deterioration, which supports step down and discharge decisions. The system ingests EHR and patient monitoring data over an HL7 connection and runs on infrastructure the hospital provides. Around the models sits a tele ICU platform with unit level situational awareness, a rounding worklist, a configurable rules based notification tool called NotifyMe, and deep links back into the medical record. Regulatory history is the deepest in this category. An Emergency Use Authorization for the respiratory deterioration model was granted in June 2020. The first 510(k), K200717, cleared in 2021 and was described by the company as the first FDA clearance for a device predicting hemodynamic instability in the ICU. A second clearance, K233216, was granted on 13 January 2024 under 21 CFR 870.2210, product code QNL, Class II. That submission did two things: it broadened the intended use environment from the ICU alone to all hospital critical care areas including emergency department resuscitation, post anesthesia care, step down and specialised units, and it demonstrated that the models met the same pre specified performance criteria after being retrained on a reduced feature set of 50 inputs rather than the original 80. FDA also cleared a Predetermined Change Control Plan permitting further retraining without a new submission, subject to stated conditions. Evidence spans the regulatory file and the peer reviewed literature. The 510(k) validation was a retrospective cohort study run independently across two datasets, the University of Massachusetts eICU dataset of 6,534 patient stays and the public MIMIC III dataset of 5,069 patient stays. A 2023 paper in CHEST compared the models against telemedicine system alerts and bedside monitor alarms at UMass Memorial and WakeMed. Named deployments include UMass Memorial, WakeMed Health and Hospitals, EQUUM Medical, Sheba Medical Center and Tel Aviv Sourasky Medical Center. The company maintains US operations out of Boston. Pricing is not published.

The Epic Sepsis Model is a proprietary sepsis prediction model built into the Epic electronic health record, scoring hospitalised patients continuously and firing an advisory to clinicians above a threshold Epic recommends setting at 6. It has been deployed at hundreds of United States hospitals, which makes it almost certainly the most widely used clinical prediction model in existence. It is also the most externally validated, and that history is the reason this record matters. Epic reported hospital level performance for the original model at an area under the curve of 0.76 to 0.83. In 2021 Wong and colleagues at Michigan Medicine published an external validation in JAMA Internal Medicine covering 38,455 hospitalisations, and found an area under the curve of 0.63, sensitivity of 33 percent, specificity of 83 percent and positive predictive value of 12 percent. An accompanying editorial was titled The Epic Sepsis Model Falls Short. A separate study of 145,885 encounters across two county emergency departments later reported sensitivity of 14.7 percent and, more strikingly, a median alert lead time of zero minutes, meaning half of all alerts arrived at or after the point clinicians had already recognised sepsis. Epic subsequently rebuilt the model. In February 2026 the same lead author published a multicentre prospective validation of Epic Sepsis Model version 2 in JAMA Network Open, covering 227,091 inpatient encounters across four major United States health systems, reporting an area under the curve between 0.82 and 0.92. Discrimination improved substantially. The paper also found high institutional variability, low positive predictive value and high alert burden, and recommended that institutions run local validation, integrate workflows to handle false positives, and adopt alert silencing strategies. This record is scoped to the sepsis model and is not an assessment of Epic Systems or its other AI features.

Clinical risk platform running real time machine learning models inside hospital EHRs to detect deteriorating patients, with early sepsis warning as the flagship use case and additional configured uses spanning clinical deterioration, pressure injuries, palliative care, and transitions of care. A Johns Hopkins spinout founded on roughly a decade of academic research, it is one of very few clinical AI vendors whose deployed system has been evaluated in large prospective multi site studies published in peer reviewed journals.