Kinometrix

The model is the product. The Kinometrix Fall Risk Assessment Solution (K-FRAS) is described as a proprietary predictive machine learning model, developed on real patient electronic health records, that reads the record in the background, computes a fall risk profile with the specific drivers of each patient's risk, and writes it back into the record with recommended interventions.

There is no hardware, no interface of its own (the company calls the platform headless) and no services layer, so removing the model leaves nothing to sell. A pressure injury risk model is described as in development.

Nothing clinical is automated. The system computes a risk profile, writes it into the record and prompts the care team with interventions matched to that profile, and nurses decide what to do. The nurse's own assessment is an input to the model rather than something it replaces, and each prediction names its drivers, so a clinician can disagree with a specific driver rather than only with a number.

No threshold, missed fall rate or abstention behavior is published, and nothing says what happens when the model's score and the nurse's documented assessment disagree, which one the chart treats as the assessment of record, or what a unit sees if the model stops updating.

The model is described only as a proprietary predictive machine learning model, built with best in class machine learning on real patient EHR data. No model class, feature list, training population, validation method, calibration, operating characteristics or version is published.

Two things are stated at a high level: the inputs are existing record data including the nurse's own documented assessment, and each prediction is returned with the specific drivers of that patient's risk, which lets a clinician see what the model is weighting.

Nothing establishes who else sits between a patient record and a score. The company says it built its own model, but no hosting provider, cloud platform, subprocessor list or processing location is published, and it is not stated whether the model runs inside the hospital's environment or the record is sent out to be scored.

No named deployment and no performance claim a reader can check. The company says K-FRAS predicts fall risk with accuracy superior to all other fall risk prediction tools, with no figure, comparison, population or method. Its white paper, by its chief clinical officer, is described as a review of three existing fall risk tools rather than an evaluation of K-FRAS, and PubMed lists no study of the product.

The only named relationship is a 2023 partnership with AvaSure, a virtual sitting company, for fall risk identification, which is a distribution partner rather than a hospital result. The company's pages were last updated in 2023.

Nothing is published on how patient data moves through the system. The company's site carries no privacy policy, terms or security statement. What the pages describe is the data flow in outline: the model reads existing record data continuously, including the nurses' own documented assessments, and writes a risk profile with its drivers back into the chart.

Where that processing happens, what the company keeps after a score is written, and whether data from customer hospitals, including their falls, is used to develop the model are all unstated. The last matters most for a fall model, which improves by learning which predictions were followed by falls, and a hospital's falls are reportable safety events.

No statement of HIPAA or business associate status and no privacy document of any kind is published on the company's site. The product reads protected health information continuously and also creates it: a risk score written into the chart is new protected health information and part of the legal medical record.

So an agreement needs to say what the company may write, into which fields, who is recorded as the author, whether a written score can be amended or withdrawn, and what happens to written content and any retained data if the relationship ends.

Nothing is published: no certification, attestation, penetration test, trust center or description of controls. That gap weighs heavily because the product holds standing access to the record in both directions, reading without a clinician invoking it and writing risk scores without one either.

The company does not describe how that credential is issued, scoped and rotated, what identity the writes carry in the record system's audit log, whether a hospital can tell a system written score from a nurse entered one, or what happens if the system cannot reach the record, since a score that silently stops updating looks the same as a current one.

No FDA clearance is claimed and no regulatory position is stated. The fall risk score comes with the drivers behind it and a nurse decides what to do, which gives a clinician a basis to review each output. That basis is thinner than what is marketed.

The model predicts which patients will fall and recommends interventions from the risk profile, and a pressure injury risk model, which predicts a condition, is in development; the company publishes no reasoning for how either sits relative to device regulation.

Nothing is published on how the model's behavior is governed or tested: no subgroup performance, calibration, fairness evaluation or responsible AI policy. Two exposures are specific to this design. A model trained on recorded inpatient falls learns from an outcome that is captured unevenly: unwitnessed falls, near misses and falls without injury are documented differently by unit, shift and staffing level, so the model can learn where falls get reported as well as where they happen. And because the nurse's own fall risk assessment is one of its inputs, variation in how nurses assess patients is carried into the score rather than corrected by it.

There is a way to question an output and nothing stands behind it. Each prediction comes with the specific drivers of that patient's risk and is written into the record, so a nurse can see what is pushing the score, disagree with a driver, and a unit can later review what was scored and when.

No error rate, missed fall rate or calibration is published, and the company says only that K-FRAS predicts fall risk with accuracy superior to all other fall risk prediction tools, with no figure, comparison or method. No terms of service, warranty, indemnity or remediation commitment is published. Because the score is part of the chart, a documented low risk score followed by a fall is a permanent record with obvious weight in any later review.

Integration is claimed with any EHR and nothing is named. The company describes the platform as headless, with a bidirectional API, reading the record in the background and writing risk profiles back wherever clinicians already view fall risk, with no extra documentation for nurses. No record system, interface standard, marketplace listing or reference site is named, so none of it can be checked before contact.

The only statement of where the system runs is that the platform integrates within a hospital's EHR system, where it analyzes patient data and sends risk profiles back. That implies processing alongside the record but commits to nothing: no hosting model, cloud or on premises option, region or residency statement is published. The site names no hospital deployment, customer count or implementation timeline either. The company's pages were last updated in 2023, when it joined the Techstars Los Angeles accelerator, and nothing more recent shows its current commercial status or reference customers.

Nothing commercial is public before a sales conversation. The product pages say to contact the company for details and pricing, and publish no price, unit of charge, band, license structure or implementation scope. The integration is described as configurable per hospital, which suggests per site setup work, but nothing says whether it is charged.

Coverage is stated clearly and nothing validates it yet. The live product assesses fall risk for hospital inpatients, and a pressure injury risk model is in development; the company frames a wider aim of addressing hospital acquired conditions generally. No unit types, patient populations, age limits or exclusions are described, and no validation of any setting is published.