Cardiology · Ambulatory ECG monitoring and AI arrhythmia classification
Zio (ZEUS System)
iRhythm Technologies, Inc.
No AI medical device has a larger published evidence engine than the Zio service: the deep network behind it was validated against cardiologists in Nature Medicine, the patch itself carries a JAMA randomized screening trial (mSToPS), a 26,751-patient diagnostic-yield cohort, and the largest comparative-effectiveness analysis in ambulatory monitoring — 287,789 Medicare patients. What keeps a genuinely global, RCT-backed platform at two marks is the post-market file: a May 2023 FDA warning letter over the Zio AT variant's undisclosed transmission limits and two late-reported patient deaths, remediated through two corrective clearances in late 2024 but without a publicly confirmed close-out.
Performance Metrics
Clinical Evidence
The evidence base spans every tier of the pyramid, which is rare in this catalogue. At the algorithm level, the Hannun and Rajpurkar Nature Medicine study (2019) validated a 34-layer convolutional network — developed on 91,232 single-lead Zio recordings from 53,549 patients — against a cardiologist committee across 12 rhythm classes, reporting an average F1 of 0.837, exceeding the averaged individual cardiologists on most classes. At the device level, Barrett and colleagues (American Journal of Medicine, 2014) had 146 patients wear a 24-hour Holter and a 14-day Zio patch simultaneously: over total wear time the patch detected 96 arrhythmia events against the Holter's 61. Turakhia and colleagues (American Journal of Cardiology, 2013) reported 99 percent median analyzable time across 26,751 consecutive patients, with more than half of first symptomatic arrhythmias occurring after the 48-hour window a Holter covers. At the outcomes level, the mSToPS randomized trial (JAMA, 2018) enrolled 1,732 at-risk adults into home-based patch screening and found new atrial fibrillation diagnosed in 6.7 percent at one year versus 2.6 percent under routine care, with downstream anticoagulation initiation in the monitored group. The CAMELOT retrospective analysis (American Heart Journal, 2024) of 287,789 Medicare beneficiaries found Zio-style long-term continuous monitoring independently associated with the highest arrhythmia-diagnosis yield and lowest retest rate of the four ambulatory monitor classes. The honest caveats: the Nature Medicine and CAMELOT work is company-funded or company-affiliated, mSToPS was an open-label siteless design, and the [[rct]] evidence speaks to detection yield, not yet to stroke or mortality reduction.
| Study | Design | n | Sensitivity | Specificity | AUC | Published |
|---|---|---|---|---|---|---|
| Hannun AY, Rajpurkar P, Haghpanahi M, et al. (Stanford / iRhythm data) | RetrospectiveRetrospective | 328 | avg F1 0.837 (algorithm) | vs 0.780 averaged cardiologists | 0.97 (class-averaged ROC) | Nat Med 2019;25:65-69 — 34-layer DNN, 12 rhythm classes; test set of 328 recordings adjudicated by a cardiologist committee; trained on 91,232 Zio ECGs from 53,549 patients |
| Steinhubl SR, et al. — mSToPS randomized clinical trial | RCTRCT | 1,732 | n/a (screening-yield endpoint) | n/a | — | JAMA 2018;320:146-155 — new AF 6.7% vs 2.6% at 1 year; siteless, open-label; anticoagulation initiated in 5.7% of monitored |
| Barrett PM, et al. — simultaneous Holter vs 14-day patch wear | ProspectiveProspective | 146 | 96 vs 61 arrhythmia events over total wear | Holter superior within the shared 24-h window | — | Am J Med 2014;127:95.e11-17 — patch preferred by patients; yield advantage comes from wear duration |
| Turakhia MP, et al. — consecutive-patient diagnostic-yield cohort | RetrospectiveRetrospective | 26,751 | arrhythmia detected in 62.2% (full wear) vs 43.9% (first 48 h) | median analyzable time 99% | — | Am J Cardiol 2013;112:520-524 — mean wear 7.6 days; over half of first symptomatic arrhythmias after 48 h |
| Reynolds MR, et al. — CAMELOT Medicare comparative effectiveness | RetrospectiveRetrospective | 287,789 | highest arrhythmia-diagnosis yield of 4 monitor classes | lowest retest likelihood | — | Am Heart J 2024;269:25-34 — Medicare fee-for-service claims 2016-2019; company-funded real-world analysis |
Clinical Pulse
Verified clinician sentiment
Verified clinician reviews launching soon · Apply to contribute →
Reviews are signed by clinicians verified against their regulator's public register.
Inside the algorithm
Editorial featureHow ZEUS reads two weeks of heartbeats.
Five stages — from raw input to verdict — drawn from manufacturer documentation and the public regulatory record.
- INGEST
- NORMALISE
- DETECT
- LOCALISE
- VERDICT
Stage 01 · INGEST
Up to 14 days of single-lead ECG arrive from the patch.
The Zio patch records one channel of ECG continuously for up to 14 days — no leads, no battery changes, showers included. In the Zio XT and Zio monitor service the full recording is uploaded after wear; in Zio AT a gateway transmits detected events during wear, within the documented per-wear transmission limits.
Wear compliance is the quiet advantage: the 26,751-patient cohort reported a median 99 percent of wear time analyzable, and over half of first symptomatic arrhythmias occurred after the 48-hour mark a conventional Holter never sees.
Input
Single-lead ECG
Inference
Per-recording
Inside the Auris+ Listing
Five more sections complete this device’s Auris+ Listing.
Decision Ledger
ProPro unlocks a private, cross-vendor log of every case you read on this device — what the AI called, what you concluded, and a one-line reason if you overrode it. Ready for the EU AI Act's deployer logging obligations when they land in 2028.
Clinical Evidence Deep Dive
ProPro unlocks the structured clinical-evidence summary — study count, target patient population, and a tabular accuracy-metrics view drawn from peer-reviewed sources.
Peer-Reviewed Publications
ProPro unlocks the curated peer-reviewed publication list with PubMed cross-links — the citation backbone of every editorial verdict.
Post-Market & Regulatory Conditions
ProPro unlocks the post-market surveillance summary, recall record, and the conditions of approval that bound real-world use.
AI Algorithm Version History
ProPro unlocks the chronological record of algorithm version changes — what changed when, drawn from manufacturer changelogs and regulatory filings.
Regulatory Approvals
Safety Record
Warning letter to iRhythm following a July-August 2022 inspection: Zio AT misbranded for failing to adequately disclose per-wear transmission limits (100 patient-triggered, 500 auto-detected events, beyond which events go untransmitted), promotion toward higher-risk populations beyond the cleared indication, quality-system violations, and late medical device reports including two patient deaths. Remediated via two corrective 510(k) clearances announced 21 and 30 October 2024 (design updates and design/labeling modifications); Zio AT remained on the US market. No formal FDA close-out had been made public as of the company's spring 2026 filings.
The Zio AT variant carries the catalogue's most instructive recent post-market record. On 25 May 2023 the FDA issued iRhythm a warning letter following a 2022 inspection, citing misbranding — physician-facing materials did not adequately disclose that a Zio AT wear can transmit at most 100 patient-triggered and 500 automatically detected arrhythmia events, after which further events go untransmitted — along with marketing toward higher-risk patient populations beyond the cleared indication, quality-system deficiencies, and failure to file medical device reports on time, including two patient deaths. iRhythm remediated through two 510(k) clearances in October 2024 covering previously made design updates and new design and labeling modifications; Zio AT remained on the US market throughout. As of the company's spring 2026 securities filings, remediation was still being reported to the FDA and no formal warning-letter close-out had been made public. The long-term continuous monitoring products (Zio XT, Zio monitor) were not the subject of the warning letter. A negative Zio report is bounded by wear time: an arrhythmia that does not occur during the wear window — or, on Zio AT, an event beyond the transmission cap — is not captured, and a [[false-negative]] for infrequent arrhythmias remains possible.
Intended Use & Indications
The Zio service pairs a wireless, water-resistant single-lead chest patch with the ZEUS System, iRhythm's deep-learned rhythm-classification software. In the long-term continuous monitoring configuration (Zio XT, and since 2023 the slimmer Zio monitor), the patient wears the patch for up to 14 days and returns it; ZEUS analyses the complete recording, and certified cardiographic technicians review the algorithm's annotated episodes before a report reaches the ordering clinician. In the Zio AT configuration, a gateway transmits patient-triggered and automatically detected arrhythmia events during wear for mobile cardiac telemetry-style review — subject to documented per-wear transmission limits that are central to the device's 2023 warning-letter history. The ZEUS deep-learning basis entered the cleared product line with K181502 (2018), which replaced the prior rule-based and classical machine-learning implementation. The service is prescription-only and does not render autonomous diagnoses: every published report passes through iRhythm's technician review, and treatment decisions remain with the clinician.