Current Health

A cleared machine learning wearable sits inside a business whose dominant value is orchestration and logistics.

The intelligence is genuine and was early. Class II clearance in 2019 covered an artificial intelligence powered wearable delivering continuous vital signs and patient specific alerts, and the company was among the first remote monitoring vendors to obtain clearance for a device described in those terms rather than as a passive sensor.

What surrounds it is mostly not inference. The platform aggregates data from third party remote monitoring devices and combines it with patient reported information, which is integration. Virtual visits, care coordination, patient onboarding and clinical workflow are software. And a substantial share of what makes acute care at home work is physical: assembling device kits, delivering them, installing them in a patient's home, supporting them and retrieving them afterwards, work now performed through a named logistics partner and previously through the acquirer's in home technician network.

The acquirer's own withdrawal analysis reinforces the point, describing acute care at home as requiring heavy human support including round the clock nursing triage, device sanitisation and patient outreach, and noting that those operations erode technology margins.

Graded C because a cleared model does real work at the centre of a product that is substantially a services and logistics operation. Ask what the models contribute beyond threshold alerting, and at what measured performance.

A clinician always decides, and the escalation question that defines this setting goes unanswered.

The model is human in the loop. The cleared wearable produces continuous vital signs and patient specific alerts which reach a clinical team, and care decisions rest with that team. Early material described the device as delivering treatment recommendations alongside alerts, which is a stronger claim than most monitoring vendors make and is not accompanied by any description of what those recommendations comprise or how they are bounded.

What is missing matters more here than in any ward based record in this index. In hospital at home the patient is alone, the responder is minutes to tens of minutes away, and deterioration that a ward would catch on a routine round must instead be caught by the platform and escalated. That makes escalation design the entire safety case, and nothing published states who monitors alerts and when, what response time is committed, how an emergency transfer is triggered, or what the vendor requires a health system to staff before going live.

No alert volume, false alarm rate or detection sensitivity is published either, so a programme cannot estimate either its workload or its miss rate.

The acquirer's exit analysis noted that this model requires round the clock nursing triage, which confirms the staffing burden is real and is borne somewhere, without saying by whom.

Ask for the escalation protocol and committed response time, who staffs monitoring, and the alert volume per patient day.

Capabilities are described and the strongest claim carries no number.

What is disclosed is the shape. A wireless wearable captures vital signs continuously, machine learning produces patient specific alerts, and the platform aggregates data from other remote monitoring devices alongside patient reported information to describe a trajectory. The device is characterised as artificial intelligence powered in its own clearance announcement rather than only in marketing.

The claim requiring scrutiny is accuracy at intensive care level. That is a strong and specific comparison, and no supporting figure, comparison study or measurement error was located to accompany it. Equivalence to intensive care monitoring is exactly the sort of assertion that should arrive with numbers, because a clinician deciding whether a patient is safe at home is relying on it.

The reference to treatment recommendations in early material is similarly unaccompanied. Whether the platform suggests interventions, and if so on what basis and with what validation, is undescribed.

No model card, architecture description, operating point or accuracy figure exists for any component, and nothing states what the aggregation layer computes across combined device and patient reported data.

One pre emptive note: further customer names cannot move this grade. Only published accuracy against a reference standard, and an operating point for alerting, will.

Ask for the accuracy study behind the intensive care comparison, and what the treatment recommendation capability actually produces.

The physical chain is partly named, which is unusual, and the technical chain is not.

The disclosure that earns the grade is logistics. A named national healthcare distributor's service handles last mile fulfilment, installation and retrieval of in home monitoring kits, and under previous ownership the acquirer's in home technician network performed setup and support. Naming the party that physically enters patients' homes and handles their devices is a meaningful supply chain disclosure and one that most vendors in this index would not make, partly because most do not have a physical chain to disclose.

The device layer is partly visible by implication rather than by statement. The platform aggregates data from multiple remote monitoring devices, so third party manufacturers are necessarily in the kit, and none is named alongside the company's own wearable.

Everything technical is undisclosed. No cloud or hosting provider, no sub processor register, no machine learning component and no manufacturing arrangement for the wearable was located.

Training provenance is unaddressed and the dataset would be substantial, since more than 70,000 patients across acute and chronic programmes generates a large corpus of continuous physiology paired with clinical outcomes, and nothing states whether it has informed model development or under what basis.

Ask for the sub processor register and hosting provider, which third party devices ship in the kit, and whether monitoring data trains models.

One of the deepest named customer rosters in this index, with no published outcome behind it.

The deployment claim is unusually strong and specific. The company states that more than one third of all patients in the United States who have experienced hospital at home were cared for through its platform, alongside more than 70,000 patients across care models. That is a market share claim rather than a customer count, and it is corroborated by a named roster spanning Mass General Brigham, Geisinger, UC Davis, Atrium Health, Baptist Health, UMass Memorial Health, Virtua Health and OSF HealthCare, with early work at Mount Sinai Brooklyn. Academic and large community systems running acute care in patients' homes on this platform is meaningful evidence that it functions in the highest risk remote setting.

What is absent is measurement. No peer reviewed publication was located, and no quantified result exists at any named site: no readmission change, no escalation rate, no length of stay effect, no safety outcome.

One distinction belongs on the record so it is not misread later. The acquirer's write down and exit reflect commercial and reimbursement conditions rather than a finding about clinical performance, and the same executives cited waiver uncertainty and slow sector adoption as the cause. A buyer should not treat the divestiture as evidence the product failed clinically, nor treat survival as evidence it succeeded.

Ask for outcome data from a named hospital at home programme, including escalation and readmission rates against the system's inpatient baseline.

A dedicated pass located no encryption statement, no retention schedule, no access control model, no deletion process and no position on whether patient data contributes to model development.

The home setting creates exposures that ward based monitoring does not. Devices operate on a patient's domestic network rather than a hospital's, connectivity is outside institutional control, and the household includes people who are not the patient. Nothing describes how the platform behaves on an unmanaged network or what a family member can see.

The physical chain of custody is the more distinctive gap. Monitoring kits are delivered, installed, used and retrieved, then presumably redeployed, which means devices holding one patient's physiological data travel to another patient's home. Sanitisation, data wiping and verification between deployments are basic requirements of that model and none is published, with a third party logistics provider now performing the handling.

The clinical data itself is high acuity. Hospital at home patients are sick enough for inpatient admission, so the record generated is a continuous physiological trace of a genuinely unwell person in their own home, alongside patient reported symptoms.

One pre emptive note: further deployment scale cannot move this grade. Only a published retention position and a documented device reprocessing procedure will.

Ask what is retained and for how long, how devices are wiped between patients, and how the platform operates on an unmanaged home network.

Two dedicated passes located no health privacy position of any kind: no compliance statement, no business associate agreement template, no execution requirement and no processing terms.

The delivery model makes the omission substantive rather than formal, because this product places third parties inside patients' homes. A logistics partner handles fulfilment, installation and retrieval of monitoring kits, and under previous ownership an in home technician workforce performed setup and support. Those personnel enter a patient's residence, handle devices carrying that patient's clinical data and interact with the patient directly. Whether they are covered under agreements, what they may access, and how their conduct is governed is unaddressed anywhere.

The device return path raises a further question that is specific to kit based home monitoring. Equipment retrieved from one patient's home and redeployed to another must be sanitised and cleared of prior data, and no published material describes that process or who is accountable for it.

Ownership transition adds a live question. The company passed from an independent business to a national retailer and back again, and nothing states what happened to patient data through those transitions or which entity now holds it.

Graded D because nothing at all was located. Ask for the agreement template, how logistics personnel are governed, the device reprocessing procedure, and what happened to data across the ownership changes.

Two dedicated passes located no security page, no external attestation, no trust centre, no penetration testing statement and no vulnerability disclosure policy.

The surface is wider than a software platform's and extends into private homes. Connected monitoring devices operate on domestic networks the vendor does not control, a cloud platform aggregates their output, and physical equipment circulates between households through a third party logistics operation. Device authentication, firmware update practice, pairing security and the handling of equipment that has left institutional custody are all first order questions and none is addressed.

The circulation model is the distinctive risk. Kits retrieved from one home and redeployed to another pass through warehousing and handling by a partner organisation, and a device that retains configuration or data between deployments is a breach path that hospital based monitoring simply does not have.

The ownership history adds a further consideration. A business that moved from independence into a national retailer and back out again will have had its security programme reshaped twice, and nothing published describes the current posture or who now operates it.

The buyer profile makes the absence surprising, since major academic health systems ran vendor security reviews to approve this platform, so a posture exists and is simply not published.

One pre emptive note: further clinical deployments cannot move this grade. Only an external attestation, or documentation available under agreement, will.

Ask whether any external security assessment exists, how devices authenticate and update, and how retrieved equipment is secured and cleared.

A genuine and early clearance for a machine learning device, with the binding regulatory constraint lying elsewhere entirely.

United States Class II clearance was granted in February 2019 for a wireless wearable explicitly described as artificial intelligence powered, covering continuous vital signs monitoring with patient specific alerts, and a further machine learning related clearance was reported in 2021. Obtaining clearance for an artificial intelligence characterised monitoring device in 2019 was early, and it places this vendor above the substantial cohort in this index that makes clinical claims with no device determination at all.

What is thin is the disclosure and the breadth. No clearance numbers or indications for use are published, the count appears to be one or two rather than the portfolios held by comparable monitoring vendors here, and no current European or United Kingdom marking was located despite Scottish origins.

The more consequential regulatory fact concerns payment rather than devices. Acute hospital at home in the United States operates under a federal waiver rather than permanent benefit structure, its extension has repeatedly been contested in budget negotiations, and roughly 400 hospitals across 142 systems in 30 states were operating under it as of mid 2025 against 20 programmes beforehand. A device clearance is permanent; the waiver is not, and the waiver is what determines whether the market exists.

Ask for the clearance list with indications, current international marking, and how the company is positioned if the waiver lapses.

A dedicated pass located no fairness testing, no subgroup performance, no calibration data, no governance framework and no external audit.

The measurement exposure is the same one this index has recorded against other continuous monitoring vendors and is unaddressed here. Continuous vital signs from a body worn device involve optical sensing, and optical measurement through skin performs differently by pigmentation, with pulse oximetry carrying a well documented record of overestimating oxygenation in patients with darker skin. A wearable claiming intensive care level accuracy across all patients invites that question directly, and another vendor in this same segment publishes an explicit skin tone validation claim, which shows the disclosure is achievable.

The home setting adds a second and less common dimension. Hospital at home eligibility depends on a patient having suitable housing, adequate connectivity and often a caregiver present, so the population able to receive this care is filtered by circumstance before any algorithm runs. A platform that improves outcomes for patients who qualify may widen the gap for those who do not, and nothing published examines who is enrolled and who is excluded.

Domestic environments also vary in ways that affect signal quality, from network reliability to household conditions, which means measurement performance plausibly tracks socioeconomic circumstance.

Ask for accuracy by skin pigmentation, and for the demographic composition of patients enrolled in hospital at home programmes on the platform against those declined.

Nothing contractual is published, in the highest stakes remote setting this index covers.

A dedicated pass located no service level agreement, no accuracy warranty, no uptime commitment, no indemnity and no remediation position.

Hospital at home is where a monitoring failure carries the gravest consequence. The patient is admitted, acutely unwell, and alone, and the monitoring platform is the substitute for the nursing observation a ward would provide. If an alert does not fire, does not arrive, or arrives at an unstaffed console, nobody is present to notice. No detection sensitivity is published, so the residual risk of that scenario cannot be quantified by a health system that is nonetheless accepting clinical responsibility for the patient.

The chain is unusually long and unusually unallocated. A device manufactured by one party is delivered and installed by a logistics partner into a home network controlled by the patient, transmitting to a vendor platform monitored by health system staff. A failure could originate at any point, and no published material describes how responsibility divides.

Availability is equally unaddressed despite being clinically load bearing rather than merely operational.

One pre emptive note: further named health systems cannot move this grade. Only contractual terms, or published detection and availability performance, will.

Ask for the availability commitment, what is warranted on alert delivery, and how liability is allocated across the device, logistics and platform chain.

Device aggregation is a designed capability and record system integration is undescribed.

The platform's stated function includes integrating remotely transmitted information from multiple remote monitoring devices with patient reported data, which is genuine multi source interoperability and is the harder half of home monitoring. A hospital at home programme uses a blood pressure cuff, a pulse oximeter, a scale, a thermometer and a wearable, frequently from different manufacturers, and normalising those into one clinical view is real integration work rather than a claim.

Deployment at major academic health systems implies clinical record integration exists in practice, since a programme of that kind cannot run with a parallel unconnected record.

What is absent is any specific. No electronic health record is named, no interface standard is described, no marketplace or validated integration listing was located, and nothing states whether monitoring data, alerts or virtual visit documentation reach the chart.

That gap has a particular consequence in this setting. A hospital at home patient is formally an inpatient, so their care must be documented to the same standard as a ward admission, and whether observations recorded at home appear in the admission record as discrete entries determines whether the programme can meet that standard without duplicate manual entry.

Ask which record systems are integrated in production, through what standard, and whether home observations post to the inpatient record.

A dedicated pass located no hosting provider, no region, no residency commitment, no tenancy model and no continuity position.

The architecture can be inferred only loosely: devices in the home, a cloud platform receiving their data, and clinician facing software. Nothing states where that platform runs or under whose infrastructure.

Two factors make the omission more consequential than usual. The company originated in Scotland and holds a United Kingdom heritage while now operating primarily from Boston, so whether any European processing or residual data location persists is a reasonable question with no published answer. And ownership has changed twice in four years, from independent business to national retailer and back, which raises the question of whether infrastructure moved with it and what happened to data during each transition.

Continuity deserves specific attention in this setting. A hospital at home patient is an admitted patient whose monitoring is the substitute for a ward, so platform unavailability is not a degraded service but an unmonitored inpatient at home. No availability commitment, redundancy description or fallback procedure was located.

The home network dependency compounds it, since connectivity is outside institutional control and nothing describes behaviour during a domestic outage.

Ask where the platform is hosted, what happened to data across the ownership transitions, the availability commitment, and how the system behaves during a home connectivity failure.

Cost is absent from every published surface, in a category where the payment question is larger than the price question.

Two dedicated passes located no pricing page, no unit of charge, no range, no tiering, no implementation fee position and no minimum commitment.

The unit is genuinely hard to infer because the offering spans software, hardware and physical service. A kit of monitoring devices is placed in a home, installed, supported and retrieved, so cost carries consumable, logistics and platform components that behave differently, and nothing indicates whether charging follows the patient episode, the enrolled patient, the programme or the enterprise.

The dominant commercial fact is reimbursement rather than price. Acute hospital at home in the United States is paid through a federal waiver whose extension has repeatedly been drawn into budget negotiations, and that uncertainty was named directly by a 400 million dollar acquirer as the reason the business scaled more slowly than expected before it exited and wrote down the asset. A health system evaluating this is underwriting a policy decision, and the vendor publishes nothing about how contracts are structured against that risk, whether terms flex if the waiver lapses, or what happens to a programme mid contract.

No return proxy is published either, despite avoided inpatient bed days being the clearest economic argument available in this category.

Ask for the unit of charge, how device kit logistics are priced, and how contracts handle waiver expiry.

Concentrated on the hardest remote setting, with a differentiated second programme.

The primary setting is acute hospital at home, meaning patients who would otherwise occupy an inpatient bed are managed in their own homes. That is the most demanding remote monitoring context there is, because acuity is high, the safety net is thin and nobody is in the room. A platform carrying that load is by implication capable of lighter chronic and post acute monitoring, which it also serves.

Oncology at home is named as a distinct programme rather than folded into general acute care, and that distinction is substantive. Delivering cancer treatment and monitoring toxicity at home involves different escalation triggers, different symptom burden and different clinical staffing from a pneumonia admission, and building it separately indicates the specialty was engineered for.

The named health system roster spans academic centres, large community systems and regional providers across multiple states, so the platform is not confined to one operating model.

What is not evidenced is the distribution. Nothing states how many programmes run acute versus chronic, what proportion of the 70,000 patients were acute admissions, or how many partnerships survived the ownership transition, which matters because the acquirer terminated its own remaining health system relationships during the wind down.

International reach is also unstated despite Scottish origins.

Ask for the current live programme count, the acute and chronic split, and which partnerships continued after divestiture.