AccurKardia

Automated interpretation is the entire product. The system performs beat by beat analysis, detects ventricular and supraventricular ectopic beats, measures heart rate, and automatically interprets thirteen rhythm classifications including atrial fibrillation, atrial flutter, and ventricular tachycardia. The company sells software rather than hardware and is device agnostic across patches, Holter monitors, and telemetry, so the algorithm is the whole of what is purchased.

The system produces fully automated interpretation intended to be reviewed within existing clinical workflows at cardiac monitoring companies, hospitals, and independent diagnostic testing facilities, rather than delivering a standalone diagnosis to a patient. The company has stated an explicit design commitment to explainable output rather than a black box approach, which is the mechanism that makes the physician review meaningful. Buyers should establish the over read requirements their accreditation and payer contracts impose, since the software's speed does not remove those obligations.

The regulatory pathway itself provides the transparency floor here, since a 510(k) submission requires performance data against a predicate, and two clearances mean two rounds of that. The company additionally states an explicit no black box position with explainable output, and publishes a stated 99 percent accuracy figure and a defined scope of thirteen rhythm classifications with an indicated population of adults aged 22 and older. Model architecture and training data provenance are not published.

Nothing identifies any party in the chain and the structure of this business means even a complete disclosure from this vendor would cover one hop of four. The platform is device agnostic by design, ingesting recordings from patches, monitors and telemetry hardware made by other companies, and returning interpretations to hospitals, monitoring services and independent diagnostic testing facilities.

A single patient recording therefore passes through a device manufacturer, a monitoring service, this interpretation layer and a reading facility, each a separate organisation with its own terms, its own retention practice and its own idea of who it answers to. Reconstructing where responsibility sits at each hop, and which party a patient could ask about their own recording, is genuinely hard from outside, and no participant publishes a map of it.

This index has recorded relabelled engines and embedded components before; a four party chain assembled per deployment is the most fragmented version of that problem so far. The training question carries particular weight for this vendor because continuous rhythm data at scale is exactly the corpus needed for the electrocardiogram derived biomarkers it has under development. Nothing states whether interpretation volume from customers feeds that pipeline or whether a customer can decline. Ask what is retained after an interpretation, whether recordings contribute to biomarker development, and which party in the chain holds the obligation to the patient.

Two FDA clearances constitute submitted performance evidence reviewed by a regulator, which is a materially higher bar than vendor assertion and is what distinguishes this from most AI in the index. Reported performance is 99 percent accuracy across the cleared rhythm classifications with results delivered in minutes rather than days. What is not published is peer reviewed prospective evaluation in deployed clinical settings or comparative accuracy against cardiologist over read, which is what would move this to an A.

Converted from Not Rated. No stewardship terms were located, and the data path crosses more organisational boundaries than most records in this index.

Nothing was retrieved on retention of waveform or interpretation data, de identification, model training use or deletion at contract end.

The boundary crossing is the distinctive feature. The platform is device agnostic by design, ingesting from patches, Holter monitors and telemetry hardware made by other companies, and returning interpretations to hospitals, monitoring companies and independent diagnostic testing facilities. A single recording can therefore pass through a device manufacturer, a monitoring service, this interpretation layer and a reading facility, each a separate organisation with its own terms. Where responsibility for the data sits at each hop, and who the patient could ask, is genuinely hard to reconstruct from outside.

The training question has particular weight for this product. Continuous rhythm data at scale is exactly the corpus needed to develop the electrocardiogram derived biomarkers the company has under breakthrough designation. Whether interpretation volume from customers feeds that pipeline, and whether a customer can decline, is unaddressed.

Ask what is retained after an interpretation, whether recordings contribute to biomarker development, and which party in the chain holds the obligation to the patient.

Converted from Not Rated. No published position was located, and the multi party chain determines what the agreement has to cover.

No business associate agreement, addendum, role statement, subcontractor flow down, breach notification timetable or review cadence was retrieved. Customers include hospitals and independent diagnostic testing facilities, which are covered entities, so business associate arrangements certainly exist in contracting. None is public.

The chain is the substantive issue. Because the platform sits between device makers, monitoring companies and interpreting facilities, the question of who contracts with whom is not obvious. A hospital may hold an agreement with a monitoring service that in turn holds one with this vendor, in which case the interpretation layer is a subcontractor and the hospital's instrument reaches it only by flow down. Or the relationship may be direct. Those allocate breach notification and liability differently, and a buyer cannot tell which applies without asking.

The patient sits furthest from all of it. Someone wearing a cardiac patch has a relationship with the ordering clinician, and the interpretation of their rhythm data is performed by a company they have never heard of, under an agreement they cannot see.

Ask which entity you contract with, whether this vendor is a direct business associate or a subcontractor, and for the flow down terms if the latter.

Converted from Not Rated. No independent security attestation was located, and the note's teaching point is the reason to read this axis carefully for any cleared device.

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

The distinction is worth stating plainly because it is the single most common confusion in this category. Clearance evaluates whether a device is safe and effective for its intended use. It examines the algorithm, the validation, the labelling and the risk analysis. It does not examine whether the vendor encrypts data at rest, controls employee access, tests its code, monitors for intrusion or would detect a breach. A cleared device from a company with no security programme is entirely possible, and clearance is not evidence either way.

This index has now drawn the line four times across imaging and diagnostics, and it holds identically here. Premarket cybersecurity documentation, where a submission includes it, examines the security of the product. It is still not an audit of the organisation.

The holding is continuous cardiac data. The platform ingests from patches, monitors and telemetry and performs beat by beat analysis, so it accumulates longitudinal rhythm data on identified patients at volume.

Ask what independent assessment exists, and do not accept the clearance as an answer to the question.

The strongest regulatory position among the vendors added in this batch, and among the clearer ones in the index. The AccurECG Analysis System is an FDA cleared Class II Software as a Medical Device, and the platform has now secured two separate 510(k) clearances, the most recent being version 2.0 under K252361 in January 2026 with a specific cleared scope of thirteen rhythm classifications and an indicated population of adults 22 and older.

Separately, the company reports FDA Breakthrough Device Designation for an investigational ECG based hyperkalemia risk assessment tool, which is a designation to expedite review rather than a clearance and should not be read as authorization for that pipeline product.

Converted from Not Rated. The prior analysis named the right variables and the device agnostic architecture makes one of them structural rather than incidental.

No governance framework, subgroup performance analysis or bias evaluation was located.

Electrocardiogram interpretation is a well documented site of subgroup variation. Waveform morphology differs systematically by sex, and reference ranges derived predominantly from one population perform differently in others. Body habitus affects signal amplitude and lead vector. Age changes baseline morphology. None of that is speculative and none of it is addressed.

The device agnostic claim is where this record differs from a single hardware vendor's algorithm. Ingesting from patches, Holter monitors and telemetry systems made by different manufacturers means the model sees recordings that differ in sampling rate, filtering, electrode design, lead configuration and noise profile before any patient variation is considered. A model that performs well on one manufacturer's signal and less well on another introduces variation that follows procurement rather than physiology, and the clinician reading the output has no way to see it. Vendor neutrality is a genuine commercial strength and it widens the validation surface considerably.

Lead placement quality compounds both. Patches applied by patients or by staff of varying training produce variable signal, and nothing describes how the system flags a recording it should not interpret.

Ask for performance by device source, by sex and by body habitus, and what triggers an uninterpretable result.

The regulatory floor applies and the vendor adds two disclosures that make its claim more contestable than most. Two clearances mean two rounds of performance data submitted against a predicate, which is external documentation of how the algorithm behaves, and the associated obligations for reporting, complaint handling and correction exist regardless of what the vendor publishes.

On top of that the company states an explicit position against black box operation with explainable output, and it defines its scope precisely: thirteen rhythm classifications and an indicated population of adults aged twenty two and older.

Stating the indicated population is a form of limitation disclosure this index credits wherever it appears, because it tells a buyer who the product is not for, and a rhythm classifier applied outside its indicated age range is doing something that was never evaluated. Naming the number of classifications does the same for scope: a rhythm outside those thirteen is not a low confidence answer, it is out of scope.

Held below the top grade because the published 99 percent accuracy figure carries no methodology, reference standard or denominator, no per class performance is given across the thirteen, and no warranty, indemnity or remediation commitment attaches. Per class matters here because rare arrhythmias are both the hardest to classify and the most consequential to miss. Ask for per class sensitivity, the reference standard behind the headline figure, and behaviour on out of scope rhythms.

Interoperability here means device rather than chart, and on that axis it is a design principle: the platform is explicitly device agnostic, ingesting data from patches, Holter monitors, event recorders, and telemetry devices without custom integration work, and is built for flexible customer integrations at enterprise scale. That matters commercially because it lets monitoring companies and OEMs adopt without re platforming. No EHR, HL7, or FHIR integration was located.

Described as a cloud based enterprise grade platform designed for scale and high ECG volumes with flexible customer integrations. Beyond the cloud characterization, no tenancy, hosting, or data residency terms are published.

Converted from Not Rated. No pricing, basis or structure is published, and the buyer set makes the unit the question.

Customers span cardiac monitoring companies, device manufacturers, hospitals and independent diagnostic testing facilities. Those are not variations on one buyer; they are different businesses with different economics. A monitoring company reselling interpretation as part of its service, a device manufacturer embedding it in a product, and a hospital consuming it directly would each expect a different commercial shape, and nothing published indicates whether the company offers one model or several.

A volume based interpretation fee would be typical for this category and is the obvious default to test. If that is the model, it scales directly with recording volume, which matters because the customers most likely to adopt automated interpretation are those with the highest volumes and the least capacity to read manually.

The embedded case raises a further question a buyer should ask about explicitly. Where interpretation is licensed to a device maker or monitoring service and reaches the clinician inside that partner's product, the clinician may not know which company's algorithm produced the reading, and the contracting organisation rather than the clinician chose it.

Ask for the pricing unit, whether it differs by buyer type, and whether the interpretation is attributed to this vendor in the output a clinician sees.

Single modality and single specialty by design: cardiac rhythm interpretation from ECG, with a cleared scope of thirteen rhythm classifications in adults 22 and older. Settings span cardiac monitoring services, device makers, hospitals, and independent diagnostic testing facilities. The stated ambition to extend ECG into a broader diagnostic biomarker, evidenced by the hyperkalemia pipeline product, is roadmap rather than current capability.