The hypothesis was that S-Patch is at least as preferred as Holter.

The primary objective of this study was to evaluate tolerability and adherence of the S-Patch compared with Holter, assessed by wear duration, completion rates, and patient-reported experience.

The secondary objective of this study was to assess device usability and to compare arrhythmia detection rates between devices in a real-world situation.

This prospective, single-center, randomized, open-label, crossover pilot study was performed between March 8, 2020, and February 28, 2023, at a tertiary hospital in Singapore. Eligible patients included older adults aged 60 years and older, attending the ED for an unwitnessed fall, syncope, or presyncope, and referred for Holter testing. Exclusion criteria included patients with known significant bradyarrhythmia, PPM, or inability to tolerate Holter monitoring.

Eligible patients were randomized into 2 groups using a block randomization method with a 1:1 ratio and were randomly allocated to either group A (Holter for 24 h followed sequentially by S-Patch for 72 h) or group B (S-Patch for 72 h followed sequentially by Holter for 24 h), with both monitors worn for up to 4 days. An events diary and a questionnaire were used to evaluate their preference between Holter and the S-Patch recorder. The visit schedule is shown in Multimedia Appendix 1, the user preference questionnaire in Multimedia Appendix 2, and the event diary in Multimedia Appendix 3.

The S-Patch is a lightweight device containing a bioprocessor that continuously collects ECG data (for up to 72 h) and transmits them in real time via a Bluetooth connection to an Android smartphone app. The initial coupling requires an internet connection that pairs the device with a precreated patient profile. The device is easily applied with only 2 standard ECG electrodes placed typically in the V2-V3 position (Figure 1). The S-Patch sends data to the phone (to be kept within 3 m) via Bluetooth continuously and does not require an internet connection. The app allows the patient to be able to add symptoms or activities either by using a precreated list or by entering their own information. At the end of the test period, the recorded information is uploaded to a cloud-based database via an internet connection and is analyzed by an algorithm, after which a report is generated and reviewed by a consultant cardiologist according to the following clinical criteria for the definition of a major abnormality:

Results from the Holter were reviewed and analyzed as per current clinical practice. The data comparison recorded is shown in Multimedia Appendix 4.

Data collection included demographic data on age, sex, race, BMI, and comorbidities, including an existing or new diagnosis of dementia or mild cognitive impairment. Cognitive screening was undertaken using the Abbreviated Mental Test (AMT), with a score of ≥7 considered normal, and functional status was assessed using the Barthel score [19] and categorized into independent (20), mild (16-19), moderate (11-15), and severe disability (≤10). The Short Physical Performance Battery (SPPB) assessment [20] was conducted, with scores ranging from 0 (worst performance) to 12 points (best performance). Falls and fear of falling were ascertained by screening questions, “Do you have fear of falling?” and “How many falls have you sustained in the last year?” Frailty was assessed using the Clinical Frailty Scale (CFS) [21], a widely used judgment-based frailty screening tool and nationally agreed community frailty screening tool [22]. Outcomes derived from hospitalization data included length of stay (LOS); ED reattendance at 7 days, 30 days, or 6 months; emergency hospital readmission at 30 days or 6 months; and inpatient mortality.

Power calculation determined that 116 patients were required to achieve 80% power to detect a difference in abnormality rates of 20%, assuming a significance level of 5%. Assuming a 20% dropout rate, the study aimed to recruit 140 patients.

Participant characteristics at baseline and secondary outcomes were examined according to their statistical distributions. Parametric continuous data were compared using 2-tailed t tests and are presented as mean (SD), while nonparametric data were analyzed with Mann-Whitney U tests and are presented as median (IQR). Categorical data comparing participant characteristics and assigned groups were evaluated using chi-square tests or Fisher exact tests and are presented as proportions. The significance level was set at α=.05, and all tests were 2-tailed. Statistical analysis was performed using Stata statistical software (version 16; StataCorp).

The study protocol was approved by the SingHealth Centralized Institutional Review Board (CIRB; ethics approval number 2019/2969) and Clinical Trial Registration 2019/2960. Informed consent was obtained from all participants, and if patients lacked capacity to make decisions, informed consent was obtained from their legally appointed representatives.

The study recruited 73 patients, 3 of whom withdrew from the study, and 70 patients completed the study interventions. Group A received Holter first, followed immediately by S-Patch, and Group B received S-Patch first, followed immediately by Holter, providing up to 4 days of continuous monitoring. Two patients received only Holter (one declined S-Patch and the other had no recording of S-Patch as data were not transferred from the Android smartphone app). Two patients refused Holter monitoring and only had S-Patch monitoring. There were 35 patients in both groups, of whom 33 in both groups completed sequential Holter and S-Patch monitoring or vice versa. The majority of patients completed cardiac monitoring during their inpatient stay. Screening and recruitment are shown in Figure 2.

Group A participants were older than Group B participants (mean age 86.9, SD 5.8 y vs 82.9, SD 7.1 y), with a similar proportion of females in both groups. There were more Chinese patients (30/35, 85.7% vs 22/35, 62.9%) and more Indian patients (4/35, 11.4% vs 2/35, 5.7%) in Group A and more Malay patients in Group B (0% vs 8/35, 22.9%; Table 1). BMI was similar in both groups, as was the prevalence of hypertension (29/35, 82.9% vs 30/35, 85.7%) and diabetes mellitus (17/35, 48.6% vs 19/35, 54.3%), but Group B had more patients with atrial fibrillation (AF; 5/35, 14.3% vs 10/35, 28.6%) and stroke (8/35, 22.9% vs 14/35, 40%), although there were no statistical differences in comorbidities between the 2 groups. Few patients were active smokers across the groups (2/35, 5.7% vs 2/35, 5.7%). Only a small proportion of patients (1/35, 2.9% vs 2/35, 5.7%) had left ventricular ejection fraction below 40%. There were no differences in the use of antiarrhythmic or dementia medications (6/35, 17.1% vs 3/35, 8.6%; Table 1).

Holter device was tolerated well in both groups and was worn for similar periods in group A vs group B (mean 22.6, SD 1.3 h vs 22.5, SD 1.1 h; P=.52; Table 1). S-Patch was also tolerated well in both groups; however, the device was worn for longer in group B, where S-Patch was administered first (mean 62.9, SD 13.4 h vs 68.8, SD 9.5 h; P=.04; Table 1). The overall S-Patch analysis time was longer in group B (mean 53.1, SD 13.4 h vs 58.9, SD 13.6 h; P=.08) but did not reach statistical significance. Although there was no difference in the patient’s device preference between groups, a higher proportion of participants in group A (3 vs 22) vs group B (7 vs 17) preferred S-Patch, with 9 participants in both groups having no preference (Table 1). Compliance with completion of preference questionnaires was high in both groups (group A, 34/35, 97% and group B, 33/35, 94%; Table 1).

Overall, 45.7% (32/70) of patients had dementia or cognitive impairment, either known or newly diagnosed during the index admission, with similar numbers of patients with cognitive impairment across both groups (group A: 18/35, 51.4% vs group B: 14/35, 40%; Table 1). Only a small proportion of patients (9/70, 12.9%) were taking anticholinesterase inhibitors (ACHi; group A: 6/35, 17.1% vs group B: 3/35, 8.6%). None of the patients on ACHi had symptomatic bradyarrhythmia. Average AMT score was 6.5 for both groups (Table 1). In terms of the duration of monitoring undertaken in those with or without cognitive impairment, both devices were tolerated well, and Holter was worn for 22.5 (SD 1.4) hours vs 22.6 (SD 1.1 hours; P=.66) and S-Patch for 67.5 (SD 9.5) hours vs 64.6 (SD 11.3 hours; P=.33; Table 2).

Frailty assessed using CFS was similar between group A and group B, with 34.3% (12/35) vs 25.7% (9/35) of participants being moderately or severely frail, 48.6% (17/35) vs 54.3% (19/35) being very mildly or mildly frail, and 17.1% (6/35) vs 20% (7/35) not being frail. There were more patients with cognitive impairment in the moderate or severely frail groups (19/35, 59.4% vs 0/35, 0%) compared to nonfrail (P=.001; Table 2). There was a trend toward more falls with increasing frailty (5/13, 38.5% in CFS 1‐3 vs 19/36, 52.8% in CFS 4 and 5 vs 14/21, 66.7% in CFS 6‐9), which did not reach statistical significance (Table 3). However, fear of falling was significantly higher in frail patients (4/13, 30.8% in CFS 1‐3 vs 14/32, 43.8% in CFS 4 and 5 vs 13/18, 72.2% in CFS 6‐9; P=.04; Table 3) but did not translate to increased hospital usage.

Both groups presented predominantly with an unwitnessed fall (21/35, 60% vs 17/35, 48.6%) and fewer with syncope (5/35, 14.3% vs 6/35, 17.1%) or near syncope (9/35, 25.7% vs 12/35, 34.3%). Falls in the preceding 12 months were similar in both group A and group B (2.2 vs 2.0, SD 1.9), with fear of falling also being similar (12/30, 40% vs 19/33, 57.6%). The SPPB was performed in 68% (24/35) of group A and 71% (25/35) of group B participants, with a mean score of 3.7 (SD 2.8) vs 3.9 (SD 3.0; P=.84; Table 1), and functional impairment was mostly cited as the reason why patients were unable to perform the SPPB. Fall frequency in the preceding 12 months was higher in those with moderate disability (3.4 falls, SD 2.9) compared with those with severe disability (1.7 falls, SD 1.3), mild disability (1.7 falls, SD 1.5), and independent patients (1.8 falls, SD 1.4; P=.03; Table 3).

In terms of arrhythmia events in group A vs group B, Holter identified 28.6% (10/35) vs 42.9% (15/35; P=.32) and S-Patch identified 9 vs 9 (25/70, 35.7% vs 18/70, 25.7%; P>.99; Table 1). There were no statistically significant differences in sinus node disease, AV nodal disease, or tachyarrhythmia between the group A and group B participants for Holter and S-Patch (Table 1). Outcome comparison between the 2 treatment arms using McNemar test showed no differences in outcome events between Holter and S-Patch monitoring, 25 out of 70 (35.7%) vs 18 out of 70 (25.7%; P=.21; Table 4). For agreement analysis using paired data from the crossover design (n=70), a 2 × 2 contingency table (Table 4) was used to compare the detection of clinically significant arrhythmias by both devices. McNemar test showed no significant difference in arrhythmia detection between the 2 devices (χ²₁=2.13, P=.21), and Cohen κ was 0.24 (P=.02), indicating fair agreement between devices. Using Holter as the reference standard, S-Patch sensitivity was 40% (10/25) and specificity was 82.2% (37/45), giving a positive predictive value of 55.6% (10/18) and a negative predictive value of 71.2% (37/52).

S-Patch identified one symptomatic arrhythmia and a pause of >3 seconds, and Holter confirmed sinus arrest, with a long pause suggestive of sinus node disease for which a PPM was recommended. Another patient had an asymptomatic 2.8-second pause on S-Patch, which was not classified as an event.

There was a high level of agreement (κ=0.8584) between the 2 blinded raters (cardiologists) for AF, which was significantly greater than expected by chance (P<.001). The observed agreement for artifact was 37.54% and fell slightly below the expected agreement of 41.18%, resulting in a low κ coefficient of 0.0582, but this was not statistically significant (P=.21).

ED attendance resulted in most patients being admitted to hospital (94.1% vs 91.4%; P>.99). Hospital LOS was similar in both groups (mean 17.9, SD 16.2 vs 16.8, SD 20.8; P=.80; Table 1). There was only 1 (2.9%, P>.99) mortality in group B during the study period (Table 1). Odds ratios for ED reattendance at 7 days, 30 days, or 6 months and for hospital readmission at 30 days or 6 months were similar between the 2 groups (Table 5).

ED reattendance within 6 months was high (5/10, 50%) in those whose Barthel score indicated severe disability, 0% (0/14) in those with moderate disability, 21.7% (5/23) in those with mild disability, and 43.5% (10/23) for independent patients (P=.007; Table 4). Those with moderate functional disability had a longer average LOS (26.6, SD 23.5 days) compared with those with severe disability (13.9, SD 9.2 days), mild disability (13.5, SD 10.4 days), and independent patients (17.1, SD 22.8 days; P=.21; Table 6).