PrecisionOS

Graded C on the mechanism, and the grade should not be read as a judgment on the product, which is among the best evidenced in this index. What PrecisionOS sells is high fidelity immersive simulation with haptic feedback, an authored library of orthopaedic procedures, and a composite Precision Score computed from named procedural parameters.

Nothing located describes machine learning: the score appears to be deterministic measurement against a validated rubric rather than learned inference, which is a legitimate and arguably more defensible architecture for competency assessment because a trainee can be told exactly which parameter cost them points. The durable asset is the validated content and the scoring rubric, not a model.

This index's workforce category explicitly grades content layer vendors down rather than excluding them, provided the mechanism is stated plainly, and this record does so. Ask the company directly which components, if any, are learned, since its marketing language around performance analytics does not distinguish measurement from inference.

The assessment instrument is better validated than anything else in this lane, and the governance around its use is undisclosed. On the credit side, the Precision Score is not a black box confidence number: it is computed from parameters relevant to safe implantation, those parameters are described in the peer reviewed literature, and the score has been correlated with real world surgical performance rather than only with itself. That is a stronger position than any competitor here.

What is missing is the human control layer. Nothing published states whether faculty can override or contextualise a score, whether a trainee can contest one, whether scores feed formal competency or promotion decisions, or who inside a programme can see an individual's history. That is precisely the disclosure this index credited Sapient AI for making and marked absent in Patient Ready, and it matters more here because the instrument is good enough to be trusted.

Transparency arrives through peer review rather than through a model card, which for an assessment instrument is the more useful route. The Precision Score's construction is described in the published literature as a composite of parameters relevant to safe and successful implantation, and the platform's validity has been characterised against the five benchmark framework simulation researchers use, meaning face, content, construct, concurrent and transfer validity, with the underlying studies available for an evaluator to read and criticise.

Held at B because the exact scoring algorithm and parameter weightings remain proprietary, no version history or recalibration policy is published, and the company's newer performance analytics and mobile metrics offerings are described in marketing language without any corresponding methodological detail.

No published statement on data handling or retention was located. Training runs on simulated cases so patient information is largely not implicated, and the sensitive object is a durable, granular, quantified record of an identified surgeon's errors. One point here inverts the usual direction of this index's reasoning and is worth stating carefully.

The company has invested heavily in validation, with randomised trials, publication across major surgical journals, and a score described as validated and correlated with real performance including transfer validity. That evidence base is genuinely strong and to the company's credit.

It is also precisely what makes this record consequential: a score with demonstrated correlation to operating room performance is a score that can reasonably be relied on by a programme director deciding whether a resident progresses, or by a department deciding who operates. The better the validity evidence, the more weight the number can bear, and the more governance the record needs. Weak evidence would make the score harmless and useless together.

Three customer types make the questions concrete and one has no education framing at all: establish the retention period, who may view an individual's score history, whether scores follow a trainee from residency into employment, and specifically what happens to performance data generated inside a device manufacturer's training programme, where the party holding it has a commercial interest in knowing which surgeons handle its implants well. At home use on personal headsets also puts some of this outside any institution's oversight.

The strongest evidence record in this category and among the strongest in the index, which is why this vendor's earlier exclusion from consideration was an error worth recording.

The core study is a block randomised, intervention controlled clinical trial of senior orthopaedic surgery residents published in JAMA Network Open, comparing immersive VR training against technical video instruction and reporting significant improvement in knowledge and procedural metrics, a transfer effectiveness ratio of 0.79, and approximately 50 percent fewer critical surgical errors. A further randomised controlled trial in the Journal of Bone and Joint Surgery addressed complex skill acquisition.

Transfer validity is the outcome that actually matters for training software, and almost nothing in this lane demonstrates it. Here it has been shown in blinded assessment of performance on real procedures rather than inferred from in simulator improvement. The company reports satisfying all five simulation validation benchmarks through multiple randomised trials in major journals. Recognition includes an American Shoulder and Elbow Surgeons top ten abstract and a gold award at Reimagine Education selected from roughly 1,500 education technology companies.

The conflicts are disclosed rather than concealed, and the design mitigates them. The JAMA Network Open paper states the chief executive's founder equity and salary and a co author's equity, records PrecisionOS as a co funder alongside the Canadian Shoulder and Elbow Society, and documents that a third party researcher observed each session with no PrecisionOS affiliate present throughout. That is the candour standard this index credits in Lyssn, with real bias mitigation attached.

No published statement on data handling or retention was located, and the framing the earlier assessment established is confirmed. Training runs on simulated cases, so patient information is largely not implicated. The sensitive object is a durable, granular, quantified record of an identified surgeon's errors.

The second pass sharpens that in a way worth stating carefully, because it inverts the usual direction of this index's reasoning. The vendor has invested heavily in validation: randomised trials, publication across major surgical journals, and a proprietary score described as validated and highly correlated with surgeon performance, with transfer validity among the benchmarks claimed. That evidence base is genuinely strong and it is to the company's credit.

It is also precisely what makes the record consequential. A score with demonstrated correlation to real operating room performance is a score that can reasonably be relied on by a programme director deciding whether a resident progresses, or by a department deciding who operates. The better the validity evidence, the more weight the number can bear, and the more governance the record needs. Weak evidence would make the score harmless and useless together.

Three customer types make the questions concrete, and one has no education framing at all. Establish the retention period, who may view an individual's score history, whether scores follow a trainee from residency into employment, and specifically what happens to performance data generated inside a device manufacturer's training programme, where the party holding it has a commercial interest in knowing which surgeons handle its implants well.

At home use on personal headsets means some of this is generated outside any institution's oversight.

No compliance statement or business associate agreement terms were located, and the earlier assessment was right that the axis maps imperfectly: simulation training on synthetic cases does not ordinarily involve patient information, so the absence is expected rather than a gap.

The conditional the earlier assessment raised remains the one to check, and it is a live possibility in this specialty. Establish whether any module builds scenarios from patient derived imaging or case data. Orthopaedic simulation adjacent products commonly do, since a realistic anatomy model has to come from somewhere, and if patient imaging underlies any scenario the analysis changes even though no identifiable patient appears to the learner.

The second pass makes clear what actually governs, and a buyer should ask against those frameworks rather than this one. In academic deployment the record is a student record, and in the United States that brings education privacy law with its own access, correction and disclosure rules. The vendor's curriculum is explicitly aligned to the graduate medical education accreditation framework and its Canadian equivalent, which means scores can feed formal competency assessment and therefore a resident's progression file. In health system deployment the same data is an employment record, governed by employment law and any collective agreement. In device manufacturer deployment it is neither, and no professional framework attaches at all.

So the right question is not whether a business associate agreement exists but which agreement governs the performance record in each of the three settings, and whether an individual can see and contest their own.

No attestation, trust centre or security page was located publicly, and the earlier assessment's reasoning holds: a platform holding identified performance data on named surgeons across residency programmes and manufacturer training makes an attestation a reasonable expectation.

The second pass adds scale and an access pattern that together define what an examination would need to cover.

Scale first. The vendor describes collaborative affiliations with more than sixty major medical institutions and deployment across more than fifty countries. An estate of that reach holds identified scores for a substantial share of a surgical generation, which is a concentration that does not exist elsewhere in this lane.

The access pattern is the more specific point. Training is no longer confined to institutional simulation labs. The vendor describes personal headsets issued for at home use around the clock, with a companion mobile application delivering performance insight. So the data path runs from a resident's own device, on their own network, to the vendor, and back to their phone. The institution whose trainee is being scored may sit outside that path entirely.

That raises questions no institutional security review would otherwise reach: how devices are enrolled and revoked, what is retained on a headset between sessions, what happens when a resident finishes a programme or a headset is lost, and whether an institution can see or restrict any of it.

The multi party structure compounds separation. Ask which report is held or scheduled, its scope, the device enrolment and revocation model, and how residency, health system and manufacturer tenants are separated.

No device authorisation and none required, since training software making no claim about an individual patient falls outside device regulation. The relevant exposure for this lane is different and this index has recorded it before as the discoverability reframe: a quantified performance score tied to a named surgeon is exactly the artifact a malpractice plaintiff would seek in discovery.

State peer review privilege statutes vary and were not drafted with externally generated, continuously produced competency scores in mind, and a score held by a commercial vendor rather than an internal quality committee may complicate a privilege claim. No position on this is published and buyers should obtain one, particularly programmes intending to retain scores as part of a formal competency record.

No subgroup or fairness disclosure was located, and there is a specific confound literature that makes the omission notable rather than routine. Published work on VR surgical dexterity has examined how posture, handedness and visual magnification affect measured performance, which establishes that scores in this modality are sensitive to factors unrelated to surgical judgment.

Left handed trainees, those with differing visual acuity or interpupillary distance, and those with prior gaming exposure or susceptibility to simulator sickness may all score differently for reasons that have nothing to do with operative competence. PrecisionOS publishes no breakdown of Precision Score distribution by handedness, sex, prior VR experience or any other characteristic.

That matters because the instrument is validated well enough to be taken seriously in assessment decisions, and an assessment instrument trusted without its confounds characterised is more consequential than one nobody relies on.

Transparency arrives through peer review rather than through a model card, which for an assessment instrument is the more useful route, because what a programme director needs to know is not how the score is computed but whether it measures what it claims to.

The score's construction is described in the published literature as a composite of parameters relevant to safe and successful implantation, and the platform's validity has been characterised against the five part framework simulation researchers use, covering face, content, construct, concurrent and transfer validity, with the underlying studies available for an evaluator to read and criticise.

Transfer validity is the demanding one and the one most simulation vendors never attempt, since it asks whether performance in the simulator carries into the operating room rather than whether the simulator feels realistic. Held below the top grade because the exact scoring algorithm and parameter weightings remain proprietary, so a trainee who disputes a score cannot see what drove it, no version history or recalibration policy is published, and the company's newer performance analytics and mobile metrics offerings are described in marketing language with no corresponding methodological detail, which means the validated instrument and the unvalidated analytics sit under one brand.

No warranty, indemnity or remediation commitment attaches. Ask which components carry the published validity evidence and which do not, the recalibration policy, and what a learner can see about their own score's derivation.

The electronic health record is not the relevant integration surface for training software, and the surface that is relevant has no published detail. For a residency programme the questions are whether Precision Scores export into a learning management system, whether they map to ACGME milestones or equivalent competency frameworks, and whether the data can leave the platform in a portable format. None is addressed publicly.

Oxford Medical Simulation, assessed in this same lane, makes competency mapping and tracking an explicit named feature; PrecisionOS does not, and a programme intending to use scores in formal assessment should confirm what it can actually get out.

Delivered through immersive VR headsets with haptic feedback, alongside a mobile application extending performance metrics off headset. No hosting provider, region, tenancy model or residency commitment is published. As with any headset dependent product, hardware platform choice and its support lifecycle sit partly outside the vendor's control and are worth establishing, along with how many headsets a programme needs to run a cohort and who supplies them.

No institutional pricing is published for the platform, and enquiry routes through the company. One disclosed commercial fact sits alongside that: a mobile application is offered free, which gives an individual surgeon a no cost entry point and is more than most enterprise training vendors provide.

Questions to settle: whether licensing is per learner, per programme, per module or per headset; whether the procedure library is bundled or purchased individually, since orthopaedic subspecialty coverage is the main axis of expansion; whether headsets are included; and whether medical device company sponsored training carries different terms from an academic programme licence, since those are different buyers with different interests in the resulting data.

Deep in one specialty rather than broad across many, which suits an assessment instrument that has to be validated procedure by procedure. Orthopaedic surgery is the focus, with the published evidence concentrated in shoulder and elbow work including reverse shoulder arthroplasty.

Buyer types are genuinely varied for this lane: residency and fellowship programmes, professional societies, and medical device companies running training on their own implants, with deployments described across North America and Europe.

The device manufacturer channel is worth a buyer's attention as a structural feature rather than a criticism, since training delivered on a manufacturer's implants serves that manufacturer's interests as well as the surgeon's, and this index has raised the same ownership neutrality question about C-SATS inside a surgical device company. Held at B for single specialty scope and because named institutional customers are described by category rather than identified.