Studies Determine Source of False Positives in AI-Equipped Implantable Cardiac Monitors

New analyses are providing clarity on the cause(s) of false-positive alerts in implantable cardiac monitors (ICMs) and the reason it has been a persistent challenge in healthcare, even in devices equipped with artificial intelligence (AI) algorithms.

While physicians experience this burden on a daily basis, the findings presented by Implicity provide new insight into the reason(s) it persists, identifying guideline-based interpretation gaps and signal-detection issues as key drivers of non-actionable alerts across modern ICM platforms.

A cross-manufacturer analysis of 2,659 rhythm episodes from 1,710 patients implanted with ICMs from Medtronic, Biotronik, Abbott, and Boston Scientific Corp., found that, despite the incorporation of AI-based detection algorithms, 32.9% of episodes were still non-actionable, with another 30.6% deemed indeterminate. Among devices without proprietary AI algorithms, 45.4% of episodes were non-actionable and 20.1% indeterminate.

To conduct the analysis, an independent expert adjudication committee applied a standardized annotation framework aligned with international electrophysiology guidelines to determine whether device-detected episodes met the diagnostic criteria for clinically meaningful arrhythmias.

The findings provide new insight into the reasons false-positive alerts persist even as device algorithms have evolved. Investigators discovered that many alerts stem from the ways in which device algorithms interpret rhythm signals relative to guideline-defined arrhythmia criteria. When those interpretations diverge from clinical definitions, benign rhythms or signal artifacts such as premature ventricular contractions or electrical noise may be labeled as clinically significant events.

The analysis also identified specific signal-detection mechanisms contributing to these alerts. Episodes labeled as cardiac “pause” events emerged as a major driver, with 46.8% ultimately determined to be false positives caused by R-wave undersensing, where the device fails to detect a heartbeat and incorrectly interprets the signal as a pause.

“False-positive alerts remain one of the biggest operational challenges in remote cardiac monitoring,” said Niraj Varma, M.D., Ph.D., professor of Medicine and consultant electrophysiologist at the Cleveland Clinic. “Every episode flagged by an implantable cardiac monitor must be reviewed by a clinician, yet even devices equipped with manufacturer AI algorithms still generate a substantial number of non-actionable alerts. When interpretation varies across device platforms and guideline definitions are not consistently applied, it becomes more difficult for physicians to quickly determine which events truly require clinical attention.”

Building on these findings, investigators conducted a second analysis to examine whether an additional AI layer could help address these persistent false-positive alerts. The study evaluated the Implicity ILR ECG Analyzer,* a cloud-based algorithm that analyzes ICM transmissions across multiple manufacturer platforms using a standardized guideline-based framework.

The results showed that Implicity’s cloud-based AI algorithm maintained very high sensitivity for detecting clinically meaningful arrhythmias—98.3% in AI-equipped devices and 94.3% in non-AI models—while filtering a substantial proportion of non-actionable alerts. Specificity reached 61.6% and 75.6% respectively, with a consistent positive predictive value of approximately 74% across both groups, demonstrating reliable diagnostic performance across different generations of implantable cardiac monitors.

“Remote monitoring only works if clinicians can trust the alerts they receive,” electrophysiologist and Implicity Co-Founder/CEO Arnaud Rosier, M.D., Ph.D., stated. “When a large share of those alerts are non-actionable, the burden is not just operational—it diverts valuable clinical time from patients who may truly need attention. Our data shows that adding a standardized, guideline-based AI layer can reduce that noise while maintaining the high sensitivity needed to detect clinically meaningful arrhythmias.”

Implicity is a digital medtech software company striving to provide the best remote care to patients with cardiac implantable electronic devices and heart failure. The company’s platform aggregates, normalizes, and standardizes data from any implantable cardiac device across all manufacturers. Implicity’s platform provides critical health information augmented by U.S. Food and Drug Administration (FDA)-cleared AI³ algorithms, enabling healthcare providers to make more informed decisions for better patient outcomes while optimizing workflows. With access to the Health Data Hub, one of the world’s largest databases of heart disease patients, Implicity can develop its AI solutions based on more robust data. The company is protecting more than 110,000 patients in over 250 U.S. and European medical facilities.

Health Data Hub is a health data platform established by the French government to combine existing health patient databases and facilitate their usage for research and development purposes.

* The version of ILR ECG Analyzer evaluated in this study (V2) is not yet cleared by the FDA. Results may not be directly applicable to the currently cleared version.