Evaluation of the accuracy of a single-lead adhesive electrocardiogram patch monitoring device (S-PATCH3-Cardio) in patients post-myocardial infarction: a pilot study

INTRODUCTION Long-term electrocardiogram (ECG) monitoring is an indispensable technology in the diagnosis and management of cardiac conditions, especially paroxysmal rhythm abnormalities. Options include monitoring by external systems or insertable loop recorders.[1] Limitations of external systems include patient discomfort and inconvenience, which together with battery life, limit the length of monitoring possible. Implantable systems monitor for longer periods, but are invasive and record only short episodes. Both approaches have latency in transmission of information and generate enormous volumes of data that are prone to false alarms due to artefact and complicated rhythm analyses.[2-4] Recent technological advances with improved signal quality, wireless data transmission and new algorithms for noise reduction make the goal of longer-term ambulatory ECG recording more feasible and may allow new applications.[5] One population that stands to benefit includes patients with a recent acute coronary syndrome (ACS) event and who are at heightened risk of arrhythmias and sudden cardiac arrest (SCA). However, the duration of monitoring is limited by resource constraints and patient tolerability. Currently, patients can only be reasonably monitored for short periods with systems that confine them to the ward. While most arrhythmias occur during the first 2–3 days after an ACS event, some are subacute and extend to post-discharge. Studies showed that up to 83% of SCA occur after discharge and are missed by the current practice of ECG monitoring only as an inpatient.[6] These events may be preventable if there is an accurate yet portable system that can monitor patients post-discharge effectively. S-PATCH3-CARDIO There is a growing market for single-lead ECG monitors, one of which was studied at our centre. The S-PATCH3-Cardio (S-PATCH3-Cardio; Samsung SDS, Seoul, South Korea) is a compact and lightweight (8 g) wireless single-lead ECG recorder comprising two standard ECG electrodes placed at the apex and sternum regions linked by a short cable. It operates wirelessly and is unobtrusive. It is powered by a standard CR2032, 3-V, coin-sized battery and can last for 100 h of continuous operation. The battery can be replaced by the patient to extend monitoring. The patch, with the accompanying Samsung SDS Cardio App, is intuitive. Initial set-up requires an Internet connection to create a patient profile and couple the S-PATCH3-Cardio with a smartphone running the Cardio App. Thereafter, the device operates wirelessly, continuously recording ECG and streaming it to the smartphone via low-energy Bluetooth at an analogue to digital conversion sampling rate of 256 samples/s. The Cardio App allows the patient to annotate symptoms or activities through a standard list or by free text, which are time stamped to the rhythms captured. This data are uploaded to a secure cloud portal via the smartphone’s mobile data connection. The uploaded ECGs are analysed by a machine learning algorithm to remove artefacts and diagnose the rhythm. Physicians can access their patients’ data remotely via the Samsung SDS Cardio Physician Web to review automatically annotated events and generate reports. STUDY DESIGN We performed a proof-of-concept study to demonstrate that such an adhesive ECG patch could be used for ambulatory monitoring of patients after an ACS event. We prospectively enrolled 42 patients admitted in an academic hospital in a 3-month period from January 2018 to March 2018. Ethics approval was obtained from the National Healthcare Group Domain Specific Review Board (Reference: 2017/00369). Stable patients above the age of 21 who were admitted for an ACS event (both ST elevation myocardial-infarction and non-ST elevation myocardial infarction) and who required telemetry for more than 24 h were approached. To minimise disruption that interventions and procedures can have on cardiac monitoring, recruitment was done only on the second day after they had undergone percutaneous coronary intervention or if no intervention was planned. Subjects were provided with a Samsung S6 smartphone with data for the duration of the study. The ECG recordings and the final report for both telemetry and S-PATCH3-Cardio were collected simultaneously for up to 48 h. Recording was terminated early if telemetry was discontinued by the clinical team or if the patient was discharged. In our institution, telemetry was acquired using a three-lead system and monitored continuously by a trained nurse and all arrhythmia events were manually evaluated and annotated. If needed, the on-call medical team was consulted. The report for S-PATCH3-Cardio was based entirely on automated analysis by the aforementioned machine learning algorithm and was not further cleaned by human analysis [Figure 1]. This was compared against conventional telemetry, looking out for critical arrhythmias such as ventricular tachycardia (VT), ventricular fibrillation and supraventricular tachycardia (SVT).Figure 1: Sample rhythm strips from the S-PATCH3-Cardio monitoring device. (a) Correctly Identified supraventricular tachycardia (SVT) strip. (b) Correctly Identified ventricular tachycardia (VT) strip. (c) Incorrectly Identified VT strip due to noise.RESULTS All patients successfully completed the recording for an average of 44.5 (interquartile range [IQR] 24.0–28.0) h with no early termination. All subjects were supplied with a fully charged S-PATCH3-Cardio, and none required a battery change across the duration of monitoring. There were no episodes of device failure or reports of intolerance of the recording devices. The signals from the S-PATCH3-Cardio were robust, with the longest duration of signal drop-off being 15.38 ± 24.38 s. Most tests had a signal drop-off within 10 s, but data were skewed due to two tests which experienced a longer maximum drop-off of 69 s and 155 s, respectively. An average of 4.59 (IQR 3.00–12.18) h of signal was removed by the software as noise. This left an average of 137,275 ± 60,996 heart beats to be analysed from the recordings over 29.66 ± 9.63 h. Based on offline analysis, S-PATCH3-Cardio performed favourably in identifying critical arrhythmias, but was more susceptible to noise. S-PATCH3-Cardio detected more events than telemetry [Table 1]. The SVT episode was correctly identified by both systems. Of the two VT episodes reported on telemetry, one was correctly identified as noise by S-PATCH3-Cardio, while the other that occurred 7 h after S-PATCH3-Cardio recording had been terminated and therefore was not captured. The rest of events picked up by S-PATCH3-Cardio were due to baseline noise.Table 1: Summary of key arrhythmias.DISCUSSION With improvements in technology, remote ECG monitoring has garnered increased interest as longer-term ambulatory and continuous monitoring has become more feasible. There are several challenges affecting the feasibility of ambulatory remote monitoring devices. Firstly, the device must be able to accurately collect an ECG tracing of reasonable quality. Next, this data must be adequately compressed and stored for efficient transfer to a storage device. Thereafter, there must be a robust algorithm to process the data to identify critical rhythms. Finally, the device form factor must be convenient in terms of portability and ease of use.[5] In pilot studies of atrial fibrillation detection where near-real-time detection is not necessary, it has been demonstrated that S-PATCH3-Cardio reasonably achieved all the above.[7] However, for arrhythmias that require faster responses, like ventricular arrhythmias, monitoring in an ambulatory setting presents added challenges. The ECG data must be automatically transmitted reliably and continuously. Subsequent analysis will require automation as continuous human monitoring of large streams of ECG data is not feasible at scale. We found that that the S-PATCH3-Cardio can consistently and reliably record, and then transmit ECG data, while the cloud-based rhythm identification algorithms are robust and compare well to in-hospital systems. Moreover, our current study found similar performance in terms of signal dropout and analysable beats as compared to these previous studies.[7] All critical events were accurately reported and no cases were missed. While there were 20 false detections of VT, this was relatively modest for monitoring 42 patients for an average of 44.5 h each, though further development of the algorithm should reduce this false-positive rate. Admittedly, the S-Patch system does not have real-time human over-reading and is unlikely to replace current inpatient telemetry monitoring systems due to its slower response time. This study was intended to demonstrate only the reliability and fidelity of ECG transmission and automated interpretation as a step towards future development as an ambulatory monitoring system for lower-risk patients. Moreover, the sample size was small and the actual events were few, and further validation in broader populations would be required. The biggest drawback any single-lead ECG system faces is the vulnerability to baseline noise.[7-10] This is a recurrent challenge regardless of the type of device, be it a handheld device or strip adhesive. Electrocardiogram is a small electrical signal and can be affected by noise from sources including myopotentials, galvanic skin currents, and electromagnetic interference. Artefact in ECG recordings often results in erroneous arrhythmia classification that may adversely affect diagnostic accuracy.[11] In our analysis, a constantly optimised machine learning algorithm was used to remove artefacts before rhythm analysis. Nonetheless, there remained several episodes wrongly identified as VT/SVT due to baseline noise. This is noteworthy as when large amounts of data are being recorded over prolonged periods, frequent false alarms generate both user and healthcare provider ‘alarm’ fatigue, rendering the system cumbersome, consequently adversely affecting effectiveness, adherence and prescription. More work needs to be done to refine the noise filtering system before it is truly fit for use in a clinical setting. Going forward, remote ECG monitoring could be extended beyond an ACS population to other high-risk populations including stroke and heart failure. This application is also noteworthy in the post-coronavirus disease 2019 landscape, where there is increasing interest in telehealth monitoring. Having a safe, reliable and convenient mode of home monitoring could facilitate earlier discharges while reducing readmissions, preserving vital hospital resources that can be redirected towards sicker patients, thereby reducing overall workload and relieving the stretched resources in our hospitals. CONCLUSION In this study, we demonstrated reasonable feasibility and accuracy of using an ECG patch recorder for the detection of arrhythmias in patients post-myocardial infarction. With further development and validation in broader populations, it could become an economical yet effective tool for the diagnosis of arrhythmias and improve preventive healthcare. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.