Accuracy data from the pilot program
This is early pilot data from 4 participating cardiology practices, reviewed against cardiologist-annotated ground truth. Not a published clinical trial. We present these numbers because transparency about what we have measured and what we have not yet measured is the foundation of clinical trust.
How we measured accuracy
The accuracy dataset consists of de-identified ECG patch records from 4 cardiology practices that participated in the ElectroKare early-access pilot. Each recording was from a commercially available 7-day or 14-day ambulatory ECG patch worn by a real patient in an outpatient monitoring program.
For each recording, a board-certified cardiologist who was independent of the ElectroKare team reviewed the full patch data and annotated arrhythmia events as the ground truth. ElectroKare's AI analysis output was then compared against those annotations across 7 classification categories.
This is a retrospective analysis on a pilot dataset. The sample reflects real-world patch quality, including artifacts, electrode lift, and noise segments. We have not published this data in a peer-reviewed journal and have not submitted for FDA clearance or CE marking. We present these numbers as an honest characterization of the system's current performance on the data we have measured, with the limitations that implies.
Performance by rhythm type
| Arrhythmia type | Sensitivity | Specificity | PPV | NPV |
|---|---|---|---|---|
| Atrial fibrillation (AFib) | 93% | 97% | 91% | 98% |
| Atrial flutter | 88% | 99% | 87% | 99% |
| SVT | 85% | 98% | 83% | 98% |
| Ventricular ectopy | 79% | 96% | 76% | 97% |
| Bradycardia | 91% | 99% | 90% | 99% |
| Tachycardia | 87% | 98% | 85% | 98% |
| Artifact (flagged segments) | 94% | 97% | 91% | 98% |
Pilot data only. Retrospective analysis against cardiologist-annotated ground truth. Not FDA cleared or CE marked. Sample from 4 cardiology practices in the early-access program.
Two cases from the pilot
De-identified clinical vignettes from the early-access program. No identifying information. Shared with permission of the participating practice.
Paroxysmal AFib detected mid-monitoring
A 67-year-old presenting for post-cardioversion monitoring wore a 14-day ambulatory ECG patch. On day 6 of the monitoring window, ElectroKare flagged a paroxysmal AFib episode lasting 22 minutes at night. An alert was routed to the responsible cardiologist. The cardiologist reviewed the ECG strip excerpt, scheduled an expedited visit, and initiated anticoagulation evaluation before the patient's next scheduled 14-day follow-up appointment.
Without automated analysis, the patch data would have been reviewed at the follow-up appointment, two weeks after the event.
Symptomatic bradycardia alert expedites evaluation
A 72-year-old with a history of sick sinus syndrome wore a 7-day patch following a complaint of recurring lightheadedness. ElectroKare flagged multiple symptomatic bradycardia events including two episodes below 38 BPM during sleep. The alert was routed as priority. The cardiologist reviewed the strip and flagged the patient for expedited pacemaker evaluation, shortening the evaluation timeline by approximately 6 weeks compared to the standard appointment queue at the practice.
The cardiologist made the clinical decision. ElectroKare surfaced the finding and made the case visible before the next scheduled appointment.
What we are and what we are not
ElectroKare is a clinical decision support tool. It is designed to surface arrhythmia findings from ambulatory ECG patch data and route them to the responsible cardiologist as a priority alert. It is not a cleared diagnostic device, and the cardiologist is responsible for all clinical decisions. ElectroKare does not replace clinical judgment.
The accuracy numbers on this page reflect our pilot dataset. They are real numbers from real patient data with independent cardiologist annotation. They are also limited: 4 practices, a single monitoring generation, no peer-reviewed journal submission as of this writing. Performance may vary on different patient populations, patch types, or signal quality conditions not represented in the pilot data.
We will publish updated accuracy data as the program grows and as we move toward formal validation studies. If you want to review the detailed methodology or discuss the data before committing to a trial, contact us at [email protected].