Biofourmis

The model is the product here in a way it is not for most records in this index.

The analytics engine builds a personalised physiological baseline for an individual patient from continuous wearable signals and raises an alert when that person departs from their own pattern. That design choice is the whole technical argument. A threshold alarm fires when a reading crosses a fixed number, which is why conventional monitoring generates so much noise: a resting heart rate that is alarming for one person is ordinary for another. Learning what normal looks like for this patient, then detecting deviation from it, is a genuine inference problem and it cannot be done with rules.

Regulatory engagement supports the claim rather than sitting beside it. A heart failure product aimed at optimising medication dosing received Breakthrough Device Designation in 2021, a status granted on the basis of preliminary clinical evidence for technologies addressing serious conditions. Very few vendors in this index have taken an algorithm down a device pathway at all, and doing so means the algorithm was described to a regulator in terms a marketing page never requires.

What holds it below the top grade is that the delivered offering is not only the model. Devices, in home service orchestration and virtual clinical teams are part of what a customer buys, and the merged business adds nursing visits by licensed clinicians. A meaningful share of the value reaching the patient is delivered by people and hardware.

Graded B: the inference is the differentiator and it does not travel alone.

Ask what the baseline model is trained on, how long it needs before it is reliable for a new patient, and what its false alarm rate looks like in production.

Humans are structurally in the loop here because the business model puts them there, which is a stronger guarantee than a policy.

The platform detects and alerts. It does not treat, prescribe or discharge. When a patient departs from their personalised baseline the system raises that to a care team, and a clinician decides whether to call, visit, adjust medication or escalate to an emergency department. The merged chronic care business goes further by employing licensed clinicians who conduct nursing visits, so the human is not a review step bolted onto an automated process but a paid part of the delivery model. A vendor whose cost base includes clinicians has a structural reason to keep them meaningfully involved.

The heart failure product is the one to watch as it develops, because medication dose optimisation is closer to a treatment recommendation than an alert, and nothing located describes what the software proposes versus what the clinician determines.

What is not described is the failure direction that matters most in this setting. A monitoring system's oversight question is not only what it does when it fires, but what happens when it does not. A patient deteriorating at home in a pattern the model does not recognise produces silence, and silence looks identical to a patient who is fine. Nothing located describes escalation timing, coverage hours, what happens when an alert is not acknowledged, or how a missed deterioration is detected after the fact.

Graded B for a genuinely human centred delivery model with the escalation guarantees unstated.

Ask for alert acknowledgement requirements, coverage hours, and how missed deterioration is reviewed.

The concept is explained clearly and the implementation is not described at all.

What is communicated well is the design idea: a personalised physiological baseline rather than a population threshold, with alerting on individual deviation. That is a genuine and non obvious architectural choice and stating it plainly is worth something.

Everything measurable is absent. A dedicated pass located no model card, no sensitivity or specificity figure for deterioration detection, no false alarm rate, no positive predictive value, no validation methodology, no description of the algorithms, no retraining cadence and no drift monitoring account.

In this category those numbers are not optional detail, they are the product. A monitoring system is characterised entirely by the trade off between missed events and false alarms, and where a vendor sits on that curve determines both clinical safety and whether the programme is operationally survivable. Too sensitive and the care team drowns and stops trusting alerts. Too specific and patients deteriorate unnoticed. Every buyer will ask this. None of it is published.

The regulatory route makes the omission more conspicuous rather than less. Clearance submissions contain performance data by construction, so for cleared algorithms the figures exist and have already been examined by a regulator. They are simply not on any public surface located.

Ask for sensitivity, specificity and false alarm rate for each cleared algorithm, the validation population, and how performance is monitored after deployment.

The physical supply chain matters as much as the model supply chain here, and neither is disclosed.

A dedicated pass located no sub processor register, no named infrastructure or cloud provider, no third party component inventory and no subcontractor list.

The hardware dependency is the distinctive one. Company material refers to clinical grade wearable devices without identifying who manufactures them, which sensor technology they use, or where they are made. That is not a procurement footnote. The sensors determine what the algorithms can detect and carry their own validated performance characteristics and their own documented limitations, so a buyer cannot assess the platform's accuracy without knowing what is measuring the patient. A device supplier changing a component, discontinuing a model or altering firmware changes the input distribution the models were built on.

Connectivity is a further unnamed link, since home devices reach the platform over some combination of cellular and consumer broadband arranged by someone, and nothing describes who provides it or what happens when it fails.

The labour chain is real and undisclosed. Virtual clinical teams, in home service orchestration and nursing visits by licensed clinicians involve people who may be employed, contracted or subcontracted, operating across jurisdictions, and nothing states which.

The merger doubles each of these questions, since two device fleets, two platforms and two service organisations were combined and no consolidated inventory is described.

Graded D.

Ask who manufactures the devices, which sensors are used, who provides connectivity, and whether clinical field staff are employed or subcontracted.

Named customers on multi year terms and a regulatory designation, with no outcome figure located behind any of it.

The customer evidence is specific and durable, which matters more than volume. A four year collaboration with Augusta University Health to expand a virtual care at home programme, a multi year agreement with Orlando Health, and named deployments at UCI Health, Lee Health and Community Health Network are all identifiable organisations committing over periods long enough to imply the programme worked well enough to renew. A partnership with a major monitoring manufacturer announced in February 2024 to extend its reach from hospital into the home is corroboration from a party with its own reputation at stake.

Breakthrough Device Designation for the heart failure product is the strongest single signal and it should be read precisely. Designation is not clearance and not approval. It is granted on preliminary clinical evidence that a technology may offer more effective treatment for a serious condition, and it buys expedited interaction with the regulator rather than a finding that the device works. That distinction is frequently blurred in vendor material and this record does not blur it.

What is absent is the outcome layer. No readmission reduction, length of stay figure, alert accuracy statistic, escalation rate or cost saving was located for any deployment, and no peer reviewed publication surfaced in this pass. For a category whose entire commercial argument is avoided admissions, the absence of a published admission figure is the gap that matters.

Ask for readmission and escalation data from a named programme with denominators, and for any published trial results.

The data collected here is denser and more intimate than anything else in this lane, and the governance around it is undescribed.

Continuous physiological monitoring produces a stream rather than a record. Heart rate, respiration, activity, posture and sleep sampled continuously for weeks describe not only a patient's clinical condition but their daily life: when they wake, whether they left the house, whether they are sleeping badly, whether they are alone. A single clinical encounter note is a snapshot. This is a behavioural record, and it is generated inside the patient's home by equipment the vendor placed there.

That makes retention the central question and it is unanswered. Nothing located states how long raw signal data is retained after a monitoring episode ends, whether it is retained at all once the patient is discharged from the programme, or what happens to it when a device is returned and reissued to another patient.

Model governance is unaddressed in the ordinary way. The engine builds personalised baselines, which by definition requires learning from individual patients, and nothing states whether anything learned from one patient informs models applied to another, or whether customer data trains or tunes any model.

The trials business adds a distinct question, since research participants consent to specific uses under a protocol, and nothing describes how those boundaries are enforced when the same platform serves care delivery.

Graded C: the collection is engineered carefully and the governance of what is collected is not published.

Ask about raw signal retention, device reissue handling, and whether learning crosses patients or customers.

The obligation is unambiguous and the evidence is absent.

A dedicated pass located no certification, no audited report, no template business associate agreement, no breach notification window, no liability cap position, no audit rights statement and no data return provision. The nearest thing to a security statement in located material is a merger era argument that consolidating vendors means a customer performs one security audit rather than several, which describes a benefit of buying rather than a control environment.

Business associate status is the correct posture for an organisation holding continuous physiological data, clinical records and trial data on behalf of hospitals, payers and biopharma.

Three features of this business raise obligations most records in this lane do not carry. Data originates in the patient's home from devices the vendor supplies, so the collection point sits outside any covered entity's premises and outside its network. Clinical staff employed or contracted by the vendor enter patients' homes and access records, which is personnel governance rather than systems governance. And the trials business handles research data under a separate consent framework from treatment data, with different rules about secondary use, yet the same underlying monitoring apparatus serves both.

Nothing located addresses any of the three.

Graded D on absence of published evidence rather than on any adverse finding.

Ask for the template agreement, the breach notification window, how clinical field staff are screened and governed, and how trial data is separated from care delivery data.

A dedicated pass located no security certification, no audited report and no trust surface of any kind.

Nothing was found covering the healthcare assurance framework, service organisation control reporting at either assurance level, information security management certification, penetration testing cadence or vulnerability disclosure. No trust centre, security page or compliance portal was located. Merger era material referred to a customer performing one security audit instead of several, which presupposes an audit without describing one.

This is recorded as absence of published evidence rather than a finding that no certification exists. A company selling into large health systems under multi year agreements would in the ordinary course hold audited reports, because those customers' procurement processes require them.

The attack surface here is broader than the usual software estate and that is what makes the silence uncomfortable. Connected medical devices in patients' homes are endpoints outside any enterprise network, communicating over consumer internet connections, physically accessible to whoever is in the house. Device security, firmware update mechanisms and provisioning are security questions in their own right, distinct from platform security, and the record is silent on both halves.

The merged estate compounds it, since two separately built platforms with separately built device fleets now sit under one company, and nothing describes whether they operate under a single security programme.

Graded D.

Ask which certifications and audited reports exist, how device firmware is updated and secured, and whether the merged platforms share one security programme.

One of the few vendors in this index that has taken an algorithm to the device regulator rather than positioning around it.

Company material describes algorithms cleared as devices, and the heart failure product received Breakthrough Device Designation in 2021. Both matter and they matter differently. Clearance means a regulator examined the intended use and the supporting evidence and permitted marketing. Designation is a status granted on preliminary clinical evidence for technologies addressing serious conditions, which expedites interaction with the regulator; it is not a finding that the device works and it is not permission to market. Vendor material across this industry routinely blurs the two, and a buyer should ask which specific products hold which specific status.

The regulatory logic is correct for what the products do. Software that detects physiological deterioration and prompts clinical intervention is making a claim about a patient's condition, which is device territory, and a product that optimises medication dosing is closer still. Pursuing the pathway rather than describing the output as wellness information is the honest route and the harder one.

What holds it below the top grade is that no clearance number, intended use statement or indication was located, so a reader cannot determine which algorithm is cleared, for what population, or with what limitations. Those documents are public once granted, and pointing to them is straightforward for a vendor that holds them.

Separate obligations attach to the trials business under research regulation and to reimbursed monitoring programmes under payment rules, and neither is addressed.

Ask which products hold clearance, under what indication, and for the clearance numbers.

This product category carries a documented, physical bias mechanism, and nothing published addresses it.

The issue is the sensors rather than the statistics. Many clinical grade wearables derive heart rate, oxygen saturation and related measures optically, by shining light through tissue and reading what returns. That measurement is affected by skin pigmentation, and the accuracy of pulse oximetry across skin tones has been the subject of substantial published concern, including findings that occult low oxygen saturation is detected less reliably in patients with darker skin. A monitoring platform inherits whatever the sensor gets wrong, and no amount of downstream modelling recovers a signal that was never accurately captured.

The consequence in this setting is silence rather than error. If a sensor systematically overestimates oxygen saturation for some patients, the platform does not raise a false alarm, it fails to raise a true one, and a patient at home deteriorates without the alert that the programme exists to generate.

A second mechanism sits above it. Personalised baselines are learned from a patient's own data during an initial period, so anything atypical about that window, including an unrepresentative baseline captured during illness, propagates into every subsequent judgement about that individual. Nothing describes how long the learning window is or how a poor baseline is detected.

A dedicated pass located no bias testing, no subgroup performance data, no statement of which sensors are used or their validation across skin tones, no fairness review and no model documentation.

Ask which sensors are used, what validation exists across skin tones, and whether alert performance has been measured by subgroup.

Recourse is undefined at every point, and the harm this product can fail to prevent is the most direct in this lane.

A dedicated pass located no indemnification position, no warranty covering algorithm performance, no accuracy guarantee, no service credit regime tied to availability or alert delivery, and no described route to dispute or escalate a failure.

The failure mode is not a bad report or a misrouted payment. A patient enrolled in an acute care at home programme has been placed there instead of a hospital bed on the understanding that they are being watched. If deterioration is not detected, or an alert is generated and not delivered, or the platform is unavailable during the hours it matters, the consequence is a patient who deteriorates at home without intervention. That is the risk the programme exists to manage and it is the risk the technology absorbs on the health system's behalf.

Responsibility in that scenario is genuinely complicated and that is precisely why it should be written down. The health system admitted the patient to the programme and holds the clinical relationship. The vendor built the detection algorithm, supplied the device, and in the merged business may employ the clinician who would have responded. A sensor may have failed, a home connection may have dropped, or a model may not have recognised the pattern. Nothing published describes how fault is allocated across those parties, and the party with the least visibility into which link failed is the one holding the clinical liability.

Graded D.

Ask what liability attaches to a missed deterioration, what the alert delivery guarantee is, and how fault is allocated between the clinical team and the vendor.

The integration story is asserted at the level of outcome rather than mechanism.

What is established is directional. The platform is sold to hospitals and health systems running care at home programmes, which is not possible without record integration, since a care team cannot run an acute patient from a separate portal alongside the chart. The merged business presented a single technology integration across the care continuum as a customer benefit, which claims integration breadth without describing it. A partnership with a major monitoring manufacturer announced in February 2024 implies device level interoperability with an installed hospital monitoring estate, which is a meaningful and difficult form of integration.

What is missing is every specific. No count of supported record systems, no named integration standard, no statement on whether the connection is bidirectional, and no description of what flows back into the chart. That last point is the one that matters most in this category. Continuous monitoring generates far more data than a record is designed to hold, so a vendor must decide what to write back: every reading, a daily summary, only alerts, or only escalations. That choice determines whether the patient's hospital record afterwards contains a usable account of what happened at home or a gap.

The home end of the chain is a second interoperability surface entirely, covering devices, connectivity in houses with poor coverage, and what happens when a patient does not wear the equipment. None of it is described.

Graded C.

Ask which record systems are supported natively, what is written back to the chart, and what happens when home connectivity fails.

A dedicated pass located no hosting model, no named infrastructure provider, no tenancy statement, no geographic residency position, no region selection, no recovery objective, no availability commitment and no failover description.

The gap is heavier here than for an analytics product because of what the system is doing. A dashboard that is unavailable for an afternoon is an inconvenience. A platform monitoring acute patients at hospital level acuity in their own homes is a safety system, and an outage means nobody is watching people who were admitted to a programme precisely because they needed watching. That is the single most important operational commitment this vendor could publish and none was located.

International operations raise residency questions the material does not answer. The company originated in Singapore, retains engineering presence in Asia and serves customers across regions, and nothing states where data is stored or who can reach it from where.

The merger adds an unresolved architectural question. Two monitoring platforms built by different companies for different acuity levels were combined under a claim of a single view across the care continuum, and nothing describes whether that has been delivered by unifying the systems or by presenting two systems behind one interface. Those are very different things for a buyer, and only one of them survives contact with an integration project.

Graded D on absence of disclosure.

Ask where data resides, what availability and recovery commitments the contract carries, what happens clinically during an outage, and whether the merged platforms share infrastructure.

Cost is absent from every published surface. A dedicated pass located no pricing page, no unit of charge, no range, no tiering, no implementation fee position and no minimum commitment.

The unit question is unusually consequential here because the offering bundles four different cost structures. Software licensing plausibly prices per patient per month or per enrolled programme. Clinical grade wearable devices are hardware with a unit cost, a replacement rate and a reprocessing question. In home service orchestration and virtual clinical teams are labour, priced by hour, visit or coverage. Nursing visits by licensed clinicians in the merged chronic care business are labour again, and the most expensive component of all. A buyer cannot tell which of those are included, which are optional, or how the mix changes the total.

Reimbursement makes the omission sharper rather than softer. Hospital at home and remote monitoring programmes are funded through specific billing pathways with their own eligibility rules and rates, so the practical question for a health system is not what the platform costs in isolation but whether the programme clears its own cost. Nothing published helps with that calculation, and the vendor is far better placed to model it than any individual customer.

The merger argument compounds it. Consolidating multiple point solutions into one vendor was presented as simplifying procurement, which is a cost argument made without a cost.

Ask what is included versus billed separately, the per patient per month figure by acuity level, device costs and replacement assumptions, and how the programme is expected to be reimbursed.

Coverage runs along the acuity axis rather than the specialty axis, which is the harder direction and the more useful one for this category.

The merged business describes an unbroken span from pre surgical optimisation through acute care at home, post acute recovery and long term chronic management. Those are genuinely different problems. Acute care at home carries hospital level acuity with the patient outside the hospital, meaning deterioration must be detected within minutes and escalation must actually arrive. Chronic management runs for years, where the objective is medication titration and slow trend detection and the tolerance for false alarms is far lower because alarm fatigue accumulates. Building for both is not a marketing claim about breadth, it is two different alerting philosophies inside one platform.

Condition coverage is stated concretely: heart failure, hypertension, diabetes and lipid management are all named, with heart failure carrying a dedicated product.

Buyer coverage extends across hospitals, health systems, payers and biopharma, with the clinical trials product serving a market that has nothing in common with the others except the monitoring apparatus underneath.

What holds it at B is evidence of depth per setting. Named deployments cluster around hospital at home programmes, and nothing located establishes scale in chronic management on the enterprise side or in the trials business. The chronic capability arrived through the merger and is described at company level rather than demonstrated at customer level.

Ask which settings the named deployments actually cover and what scale the chronic and trials businesses operate at.