Hinge Health

Deliberate B, and the reasoning is the substance of this record. The computer vision is real and deep: the company acquired wrnch, a leading human motion computer vision platform, stated it thereby held the largest computer vision team in digital health, and built TrueMotion to detect 87 body reference points from an ordinary phone camera with no specialised hardware. That is a genuine technical asset, not a label.

But the business is a digital MSK clinic: revenue comes from delivering care through employed physicians, physical therapists, nurses and health coaches, with the technology making that care team more effective and scalable. The AI is a differentiating layer on a care delivery organisation rather than the product being sold.

Not C, because unlike the rejected services businesses the technology is proprietary, substantial and central to the clinical claim; not A, because remove the AI and a care business remains, whereas removing the model from Elucid or Lyssn leaves nothing.

AI is explicitly positioned as complementing rather than replacing clinician judgement, with technology surfacing timely insights while physical therapists set and adjust care plans, and the company has published AI care principles that include continuous governance and evaluation of AI systems as clinical evidence and regulatory standards evolve. Publishing principles at all is more than most peers do.

Graded B rather than A because Robin, the AI care assistant, interacts directly with members providing instant support, and no escalation logic, containment rate or boundary description was located for what Robin handles autonomously versus what it routes to a clinician. In a population presenting with pain, where a small fraction of presentations are red flags requiring urgent assessment, that escalation boundary is the safety-critical disclosure and it is absent.

More specific than most consumer-facing digital health vendors. The provenance is traceable to the wrnch acquisition, the mechanism is named concretely at 87 body reference points tracked via phone camera, and the measured outputs are enumerated as joint angles, symmetry and endurance.

Graded B rather than A because HingeScore, the composite metric now central to the outcome claim, has no published derivation, weighting or validation methodology located, and it blends objective motion capture with subjective questionnaire responses in an undisclosed ratio. A proprietary composite score presented as an outcome measure needs its construction published.

The technology's provenance is traceable and the operating chain is not. The motion capture capability descends from a named acquisition, the mechanism is specified concretely at a set number of body reference points tracked through a phone camera, and the measured outputs are enumerated, so a buyer can establish where the core capability came from and what it produces. Published artificial intelligence care principles sit alongside, including continuous governance commitments.

What is missing is who operates it and what happens to the input. No model or model family, no hosting arrangement and no sub processor list was located, and the unanswered question about the content is more consequential here than the missing supplier names.

The product captures video of members exercising in their own homes, and nothing published states whether that video is retained, for how long, whether it is used to train models, or even whether it is processed on the device or in the cloud, which is the first thing that determines who could hold it.

Home video of a person's body is among the most intimate data any vendor in this index collects, and it is gathered from large numbers of employer sponsored members who may reasonably perceive participation as workplace adjacent. Ask whether inference runs on the device, what video is retained and for how long, whether it trains models, and for a sub processor list.

Substantial and peer reviewed, but measuring the programme rather than the AI, which is the distinction that caps this grade. The company reports three randomised controlled trials, a large-scale study of around 10,000 participants, a 12 month outcomes study published in BMC Musculoskeletal Disorders reporting pain reduced by more than half and sustained at one year, and what it describes as the only study of a digital MSK programme in an older adult population.

That is a stronger evidence base than most care delivery businesses assemble. Two honest limits keep it at B. First, the trials evaluate the overall care programme, exercise therapy plus clinician contact plus coaching, so they do not isolate any contribution from the computer vision layer, and no validation of TrueMotion measurement accuracy against laboratory motion capture or goniometry was located.

Second, no head-to-head trial exists against Sword Health or other digital MSK competitors, a gap third party sources state explicitly, so relative effectiveness in this category is genuinely unestablished.

The published AI care principles are a genuine positive and include continuous governance commitments. But the data type here is unusually sensitive and the disclosure does not match it: the product captures video of members exercising in their own homes, and no published position was located on whether that video is retained, for how long, whether it is used to train models, or whether it is processed on device or in the cloud.

Home video of a person's body is among the most intimate data any vendor in this index collects, and it is being gathered from millions of employer-sponsored members who may perceive participation as workplace-adjacent. The absence of an explicit published video retention and training-use policy is a material gap.

Converted from Not Rated, and the prior note identifies a dual status this pass has now seen twice in the same category.

No business associate agreement, addendum, role statement, subcontractor flow down, breach notification timetable or review cadence was retrieved.

The status is genuinely two sided. Selling to self insured employers and health plans, the purchasing plan is a covered entity and this vendor processes member health information on its behalf, making it a business associate. But the company also operates as a care delivery organisation, employing the clinicians who assess members and direct their programmes. An organisation providing health care through licensed clinicians and transmitting health information in standard transactions is a covered entity in its own right, not merely a business associate. The same question arises for the direct competitor graded alongside it in this pass, which is not a coincidence: it is inherent to the virtual care model, where the vendor is simultaneously a supplier to the plan and a provider to the member.

Which capacity applies determines real obligations, including whether members are owed a notice of privacy practices directly rather than through the plan.

A further layer sits alongside it: hardware components cleared as devices mean device reporting duties may also attach.

Ask which capacity the company contracts in, and whether members receive a notice of privacy practices.

Converted from Not Rated. No independent attestation was located, and the corporate status makes the absence more surprising than usual.

No SOC 2 of either type, no HITRUST, no ISO 27001, no trust centre and no penetration testing statement was retrieved.

This is a company listed on a United States exchange selling to large employers and health plans. Both facts cut against the absence. Enterprise benefits buyers run structured vendor security assessments as a gating step, and large self insured employers are among the more demanding assessors because a breach of member health data lands on the employer's own workforce. A listed company also carries securities disclosure obligations covering cybersecurity risk management, so a buyer has an avenue for information that private peers do not offer, and this index has repeatedly found filings more informative than marketing.

That is the practical recommendation here rather than a bare gap: read the annual report's cybersecurity section directly, since it is a legally attested statement about risk management and incident processes, and it exists whether or not a trust centre does.

The holding is substantial. Member musculoskeletal history, motion and visual data from home sessions, and clinical assessments across a reported large membership.

Ask for the attestation and period, and read the cybersecurity disclosure in the most recent annual filing before the sales conversation.

A split position that must not be conflated, and the distinction runs the opposite way to most index vendors. Enso, the wearable electrical nerve stimulation device, is FDA cleared. The AI is not: TrueMotion, Movement Analysis, HingeScore and Robin are not represented as FDA cleared or regulated, and they are the components this vendor is indexed for. So the company holds a real clearance, but on the hardware rather than on the software the index grades.

Graded C rather than Not Rated because, unlike Eleos or Lyssn which sit cleanly outside device regulation, a computer vision system producing joint angle measurements and a composite joint health score used to inform treatment decisions sits closer to the Software as a Medical Device boundary, and no located material addresses where the company considers that line to fall.

Published AI care principles including continuous governance are more than most peers offer, but no demographic subgroup performance analysis was located and the exposure for this technology is specific and well documented. Pose estimation and computer vision body tracking have known performance variation across skin tone, body size, clothing, mobility aids and home lighting conditions.

A member whose movement is tracked less accurately receives worse feedback and a less accurate HingeScore, and the populations most likely to be affected substantially overlap with those already experiencing worse musculoskeletal outcomes. Nothing located addresses whether TrueMotion accuracy has been evaluated across these strata. Compare Cognivue's FOCUS study as the standard this axis should be graded against.

Two passes located no warranty, indemnity or remediation commitment, and the specific gap is in the metric the entire commercial claim rests on. The composite score now central to the outcome case has no published derivation, weighting or validation, and it blends objective motion capture with subjective questionnaire responses in an undisclosed ratio.

A proprietary composite presented as an outcome measure needs its construction published, because without it nobody outside the company can tell whether an improvement reflects better function or a shift in how members answer questions, and those are different products. The commercial context makes that consequential rather than academic.

This is sold to employers and health plans on demonstrated outcomes, so the undisclosed metric is the basis on which a purchaser decides the programme works and on which further spend is justified, while the member whose data produces it has no visibility into how their score was formed.

The motion capture side is better documented, with the technology's provenance traceable to an acquisition, the mechanism named at a specific number of tracked body reference points through a phone camera, and joint angles, symmetry and endurance enumerated as measured outputs, so the inputs are inspectable even though the composite is not. Ask for the derivation and weighting of the composite, its validation, and what the vendor commits to when a reported outcome cannot be reproduced.

Converted from Not Rated. The prior note rescopes the axis correctly for this business model and finds little against the rescoped version.

There is usually no electronic health record on the buyer's side, because the buyer is an employer or a health plan rather than a provider organisation. The systems that matter are eligibility and enrolment files, benefits administration platforms, and claims feeds used to measure outcomes and avoided cost. Grading against clinical record integration would penalise the model rather than describe it, which this index has ruled against.

Against the correct surface, nothing specific was retrieved: no named benefits platform, no eligibility integration, no claims data exchange description, no standards support and no interface documentation.

The clinical continuity gap is the more interesting omission and it is real. A member's assessment, programme, progress and outcomes live with this vendor. Their orthopaedic surgeon, primary care physician or physiotherapist has no described route to see any of it. For a musculoskeletal programme that frequently runs in parallel with, or as an alternative to, conventional care, the absence of any documented path back into the member's own medical record means the treating clinician is working without it.

That is a category wide pattern rather than a failing unique here, and it is worth naming because employers buying these programmes rarely ask about it.

Ask how eligibility and outcomes flow, and whether anything reaches the member's treating clinician.

The strongest axis on this record and a real technical achievement. TrueMotion runs on the camera in a member's existing personal device with no specialised equipment, sensors or hardware required, tracking 87 body points in three dimensions. Removing the hardware dependency is what makes home-based objective movement assessment viable at population scale, and it contrasts directly with Cognivue in this same batch, which requires a dedicated calibrated device. Wearable hardware exists as an adjunct via Enso but is not required for the motion tracking. No data residency disclosure located, which is the only thing keeping this from being unqualified.

No pricing published. The product is sold to self-insured employers and health plans as a benefit, so members do not see a price and the buyer terms are negotiated and confidential. Third party comparison notes that competitors such as Kaia Health publish flat consumer subscription pricing while Hinge's cost depends entirely on employer or plan arrangement. For an NYSE listed company, contract economics are partially inferable from public filings, but nothing about per-member pricing or the fee structure is directly disclosed.

Broad within musculoskeletal care and expanding: acute and chronic MSK conditions, pre and post surgical pathways, pelvic health, and integration of behavioural health support reflecting the biopsychosocial reality that physical pain and psychological distress are interconnected.

TrueMotion specifically extends tracking to difficult regions including head, neck and hands, and the company reports roughly 75 percent of members experience pain in more than one body area, which the whole-body tracking addresses. Reach is very large across employers and health plans. Graded B rather than A because it remains one clinical domain delivered through one channel, employer and plan sponsored benefits, with no provider or direct clinical channel located.