Feasibility and diagnostic agreement of Apple Watch and KardiaMobile electrocardiograms compared with standard 12-lead ECG in children Authors: Stefan Kurath-Koller, Daniel Scherr, Nathalie Oeffl, Thomas Suppan, Hannes Sallmon, Vito Mustapic, Martin Manninger Published: 2025 ABSTRACT Mobile electrocardiogram (ECG) technologies such as the Apple Watch (AW) and KardiaMobile 6 L (KM) are increasingly used in adult cardiology, yet their performance in pediatric populations remains underexplored. This study aimed to evaluate the feasibility, tracing quality, and diagnostic agreement of AW and KM compared to standard 12-lead ECG (S12L) in children. We conducted a prospective observational study in children aged 4–17 years undergoing routine ECG evaluation. All 12 leads were reconstructed using alternative device positioning for both AW and KM. ECGs were independently analyzed by blinded pediatric cardiologists for standard parameters, rhythm interpretation, and tracing quality. A total of 30 patients were enrolled. Median acquisition times differed significantly across devices: 4 min (2;10 min; S12L), 5 min (4;10 min; KM), and 15 min (10;32 min; AW). KM yielded high tracing quality (median 9/10), comparable to standard ECG, while AW showed lower quality (median 6/10). Strong to very strong correlations (r = 0.6–0.96, p < 0.001) with standard ECGs were found for QRS axis, R and S wave voltages, and QT intervals. Both devices detected arrhythmic events. Mobile ECG devices showed strong agreement with standard ECG for selected parameters (e.g., QRS axis, major voltages), while other intervals demonstrated weaker correlation. Findings should be interpreted within the context of a small sample size and feasibility-oriented study. KM demonstrated higher signal quality and faster acquisition, while AW was limited by longer acquisition time and artifact sensitivity. Mobile ECG devices offer promising tools for pediatric arrhythmia diagnostics, although pediatric-specific adaptations are warranted. Introduction The interpretation of electrocardiograms (ECGs) remains a cornerstone in the diagnosis and management of pediatric arrhythmias. Timely and accurate ECG analysis is particularly critical in children, where symptom presentation may be non-specific, and arrhythmias often present intermittently or transiently. While conventional 12-lead ECG (standard 12-lead - S12L) remains the gold standard, its accessibility outside clinical settings is limited by logistical and technical constraints. As healthcare shifts toward decentralized, patient-centered models, interest has grown in mobile ECG technologies capable of extending diagnostic reach beyond traditional care environments1–3. Wearable and handheld ECG devices, such as the Apple Watch (Apple Inc., Cupertino, CA, USA) and the KardiaMobile 6 L (AliveCor Inc., Mountain View, CA, USA), have gained widespread use among adults and increasingly among younger populations. These tools offer advantages in terms of portability, user-friendliness, and real-time data transmission. In adult cardiology, both devices have demonstrated reasonable diagnostic accuracy in arrhythmia detection, including atrial fibrillation, premature ventricular contractions, and ST-segment changes4–6. However, their utility in pediatric populations remains largely unexplored. Children present unique challenges - including smaller body size, variable cooperation, and age-specific ECG norms - which may limit the performance and feasibility of current mobile ECG systems7,8. To date, very few studies have attempted a systematic comparison of mobile ECG tracings with S12L in pediatric patients, and none have successfully reconstructed all 12 leads using the Apple Watch (AW) or KardiaMobile 6 L (KM) in this population7,9,10. Given the increasing prevalence of mobile health technologies and their adoption by pediatric and adolescent patients, a rigorous evaluation of these tools is both timely and essential. In this prospective study, we aimed to assess the feasibility, diagnostic accuracy, and agreement of ECG parameters obtained using the AW and KM compared with S12L in healthy children. We systematically acquired all 12 ECG leads from each mobile device through alternative positioning strategies, allowing for direct comparison across modalities. By analyzing waveform fidelity, acquisition time, tracing quality, and rhythm interpretability, we sought to determine whether these mobile tools can support reliable ECG assessment in pediatric cardiology and potentially extend diagnostic capabilities into outpatient and home-based settings. Methods Study design and participants We conducted a prospective observational study to evaluate the agreement and diagnostic performance of the AW and KM compared with S12L ECG in pediatric patients. Eligible participants were children and adolescents between 4 and 17 years of age without known cardiac conditions or arrhythmias, undergoing clinical, medically indicated ECG monitoring at the Medical University Graz Children’s Hospital. Exclusion criteria comprised known cardiac disease, congenital heart disease, arrhythmias, major chest deformities, lack of compliance, or lack of informed consent. Enrollment was offered to patients admitted to the outpatient clinic of the Medical University Graz Children’s Hospital. Ethics approval and consent to participate This study was approved by the Institutional Review Board of the Medical University of Graz, Austria (EK 34–293 ex 21/22). All research was performed in accordance with the relevant guidelines and regulations, including the Declaration of Helsinki. Written informed consent was obtained from all participants and/or their legal guardians prior to study enrollment. Assent was additionally obtained from minors when appropriate. Device selection and setup We utilized the Apple Watch Series 6 (Apple Inc., Cupertino, CA, USA) and the KardiaMobile 6 L device (AliveCor Inc., Mountain View, CA, USA) for this study. Both devices were selected based on their commercial availability and widespread use among patients1. The study was independently conducted without industry sponsorship or device donation. For comparison, S12L ECG tracings, considered the „gold standard“, were recorded using the KISS ECG recording system (GE Healthcare, Chicago, IL, USA) used in clinical practice. ECG data from the AW, KM and S12L devices were pseudonymized and securely stored on dedicated study smartphones and laptop computers managed by the research team. ECG acquisition and analysis If compliance allowed, participants would hold on to the AW or KM device by themselves. Otherwise, the accompanying parent was encouraged and taught to place the devices for ECG acquisition. Time it took to acquire each ECG lead was tracked for each device. Instruction times for parents and probands were excluded. To enable valid comparison across modalities, each lead recorded using the AW and KM was mapped to its anatomically and electrically corresponding lead in the standard 12-lead ECG system. For both mobile ECG devices, all 12 leads were acquired sequentially using established positioning strategies adapted for pediatric use5. To ensure consistency, all recordings were obtained by the same investigator, who performed all device placements after standardized training. Repeated measures were not performed, in keeping with the feasibility-oriented design of the study. Heart rate values were interpreted using established age-specific pediatric reference ranges11. Accordingly, the terms ‘low heart rate’ and ‘high heart rate’ refer to values below or above the normal range expected for the respective age group, rather than fixed absolute thresholds. Apple Watch (AW) For AW recordings, the device crown was consistently used as the right-arm electrode. For the AW, sequential recordings were obtained by repositioning the device to simulate standard lead vectors, following previously validated positioning protocols for limb and precordial leads5. Limb lead I was obtained by holding the watch at the left wrist, limb lead II by placing the watch at the left leg, and limb lead III by placing the watch at the right leg while touching the crown with the left hand. Leads aVR, aVL, and aVF were acquired by established positioning techniques previously described in adult validation studies, adapted to pediatric anatomy: aVL: crown touched with the right hand while the right elbow contacted the right leg, placing the watch at the left shoulder;aVR: crown touched with the left hand while the left elbow contacted the left leg, positioning the watch at the right shoulder;aVF: crown touched with the right hand while the left hand held the right wrist, placing the watch at the left leg. aVL: crown touched with the right hand while the right elbow contacted the right leg, placing the watch at the left shoulder; aVL: crown touched with the right hand while the right elbow contacted the right leg, placing the watch at the left shoulder; aVR: crown touched with the left hand while the left elbow contacted the left leg, positioning the watch at the right shoulder; aVR: crown touched with the left hand while the left elbow contacted the left leg, positioning the watch at the right shoulder; aVF: crown touched with the right hand while the left hand held the right wrist, placing the watch at the left leg. aVF: crown touched with the right hand while the left hand held the right wrist, placing the watch at the left leg. Precordial leads V1–V6 were acquired by placing the watch at standard chest lead positions while the patient touched the crown with the right hand. All AW recordings consisted of uninterrupted 30-second segments, as required by the device. KardiaMobile 6 L (KM) Limb leads I, II, III, aVR, aVL, and aVF were recorded using the manufacturer’s recommended configuration: the right arm, left arm, and left leg were placed on the corresponding electrodes. Precordial leads V1–V6 were obtained in single-lead mode by placing the right arm on the right-arm electrode while positioning the left-arm electrode sequentially at standard pediatric chest lead locations (V1–V6). Each precordial lead was recorded separately. Each mobile-derived lead was labeled according to its intended correspondence (e.g., AW-V1, KM-aVL) and subsequently matched to the same lead from the standard ECG for paired analysis of quantitative parameters. This approach allowed for direct statistical comparison (Pearson correlation, Bland-Altman analysis) between mobile and standard ECG readings for each lead and parameter of interest. Figure 1 provides an overview on how the watch was positioned and held. Every lead had to be recorded sequentially, and a continuous recording of 30 s was obtained for every lead as required by the Apple Watch. Standard 12 lead ECG: S12L ECGs were taken within 30 min of mobile ECG recording. Standard electrode placements were applied. Fig. 1Graphical abstract illustrating methods and depicting AW and KM positioning. Leg.: LA = left arm; LL = left leg; RA = right arm; an anterior and posterior torso view showing standard 12-lead ECG electrode locations. Colored dots indicate the positions of individual precordial and limb leads (V₁–V₆, RA, LA, RL, LL) on the figurine. Color coding corresponds to the respective lead labels for visual clarity. Graphical abstract illustrating methods and depicting AW and KM positioning. Leg.: LA = left arm; LL = left leg; RA = right arm; an anterior and posterior torso view showing standard 12-lead ECG electrode locations. Colored dots indicate the positions of individual precordial and limb leads (V₁–V₆, RA, LA, RL, LL) on the figurine. Color coding corresponds to the respective lead labels for visual clarity. All ECG tracings were independently assessed by three blinded reviewers: two cardiac electrophysiologists and one pediatric cardiologist. Reviewers were blinded to device type, lead label, patient identity, and to each other’s measurements. Each reviewer evaluated all recordings for tracing quality (3-point Likert scale) and performed quantitative measurements of the predefined ECG parameters. Interobserver variability was assessed across all three readers and incorporated into the statistical analysis. All recordings were analyzed manually with regard to heart rate, QRS axis, rhythm, P/QRS/T wave duration and voltage, QRS morphology, PR and QT duration, QTc, ST segment elevation, repolarization disturbances, presence and type of arrhythmias, R/S voltage. Heart rate values were calculated as the average heart rate over the duration of each individual ECG recording. For standard 12-lead ECG and KardiaMobile 6 L recordings, which acquire all leads simultaneously, this represents a single physiological time point. For Apple Watch recordings, leads were acquired sequentially in separate 30-second segments; therefore, average heart rate values may differ between individual leads due to normal physiological heart rate variability over time. QTc intervals were calculated using standard correction formulas across the observed heart rate range. Although QTc assessment is ideally performed at heart rates between 60 and 100 beats per minute, this range is often not achievable in pediatric patients, particularly in younger age groups. QTc measurements outside this heart rate range were therefore included to reflect real-world pediatric clinical practice and were interpreted with appropriate caution. Finally, the investigator rated every lead for tracing quality and interpretability based on a 3-point Likert scale (3 points = good quality, 2 points = moderate quality, 1 point = poor quality; see Table 1 for details). Table 13-point likert scale for tracing quality assessment.ScoreDescription3 (Good)Baseline stable, low noise, clear P, QRS, and T waves. Usable for interpretation.2 (Moderate)Mild baseline drift or noise. Some components blurred but still interpretable.1 (Poor)Severe noise, baseline wander, indistinct waveforms. Not interpretable. Table 13-point likert scale for tracing quality assessment.ScoreDescription3 (Good)Baseline stable, low noise, clear P, QRS, and T waves. Usable for interpretation.2 (Moderate)Mild baseline drift or noise. Some components blurred but still interpretable.1 (Poor)Severe noise, baseline wander, indistinct waveforms. Not interpretable. 3-point likert scale for tracing quality assessment. Statistical analysis Descriptive statistics were used to summarize patient demographics, ECG acquisition times, and tracing quality scores. Continuous variables were reported as medians with interquartile ranges (IQRs), and categorical variables as counts and percentages. We used Pearson correlation coefficients to assess the agreement between ECG parameters obtained from mobile devices and the S12L ECG, as this method evaluates the strength of linear relationships between continuous variables. Prior to analysis, variables were assessed for normality using the Shapiro-Wilk test and visual inspection of histograms and Q-Q plots. The majority of key ECG parameters, including QRS axis, wave amplitudes, and interval durations, demonstrated approximate normal distribution, supporting the use of Pearson correlation. Given the clinical relevance of preserving linear measurement relationships (e.g., for voltage and time intervals), Pearson correlation was deemed more appropriate than Spearman rank correlation for this dataset. Pearson correlation coefficients (r) were calculated for each lead and for selected waveform characteristics, including QRS axis, P wave duration, QRS duration, QT interval, and individual R and S wave voltages. Strength of correlation was interpreted using standard thresholds (e.g., r ≥ 0.80 as very strong). Bland-Altman plots were constructed to visualize the agreement and systematic bias between mobile devices and standard ECG measurements across key parameters. Analyses were performed using SPSS Statistics v29 (IBM Corp., Armonk, NY, USA). Results Overall we included 30 participants (13 female) aged 4 to 17 years (median 11.5 years). Table 2 provides participant demographics, Table 3 provides an overview on specific findings. Table 2Participant demographics.Demographic dataAge (years)11.5 (4–17)Sex13 female (43%)Body weight (kg)47.8 (17–92)Body Lengths (m)1.6 (1.1–1.8)BMI18.4 (11-6-27.5)Arrhythmias2 PVCs (6%) Table 2Participant demographics.Demographic dataAge (years)11.5 (4–17)Sex13 female (43%)Body weight (kg)47.8 (17–92)Body Lengths (m)1.6 (1.1–1.8)BMI18.4 (11-6-27.5)Arrhythmias2 PVCs (6%) Participant demographics. Table 3Specific findings, including time to acquisition, quality scoring and parameter assessment of mobile devices compared to standard 12 lead ECG recordings. Correlation of Apple watch and KardiaMobile 6 L with standard 12 lead ECG using pearsons’ correlation coefficient.S12LApple WatchKardiaMobile 6 LICCAcquisition Characteristics Time to acquisition4 (2–10)15 (10–32)5 (4–10) Quality (3-point Likert scale)3 (2–3)2 (1–3)3 (2–3)Fleiss´κ 0.71Heart Rate & Cardiac Axis HRGold Standardr = 0.32, p < 0.001r = 0.44, p < 0.0010.67 QRS axisGold Standardr = 0.95, p < 0.001r = 0.96, p < 0.0010.93Intervals PR intervalGold Standardr = 0.51, p < 0.001r = 0.36, p < 0.0010.62 QRS durationGold Standardr = 0.22, p = 0.02r = 0.2, p = 0.130.58 QT intervalGold Standardr = 0.29, p < 0.001r = 0.64, p < 0.0010.81 QTcGold Standardr = 0.28, p < 0.001r = 0.47, p < 0.0010.55 R peak timeGold Standardr = 0.4, p < 0.001r = 0.53, p < 0.0010.6Voltages P wave voltageGold Standardr = 0.33, p < 0.001r = 0.49, p < 0.0010.57 R voltageGold Standardr = 0.74, p < 0.001r = 0.85, p < 0.0010.88 S voltageGold Standardr = 0.67, p < 0.001r = 0.76, p < 0.0010.84 T wave voltageGold Standardr = 0.65, p < 0.001r = 0.75, p < 0.0010.78Waveform Durations P wave durationGold Standardr = −0.04, p = 0.7r = 0.13, p = 0.330.41 T wave durationGold Standardr = 0.3, p < 0.001r = 0.26, p = 0.040.69Leg.: parameters for time to acquisition and tracing quality are given as (min.; median (range)). BMI = Body Mass Index, S12L = standard 12-lead ECG, PVCs = premature ventricular contractions, kg = kilogram, m = meters, r = Pearson correlation coefficient; values of time to acquisition and quality score are given as median (range); ICC = Intraclass Correlation Coefficient, two-way random-effects model, absolute agreement. Fleiss’ κ for categorical agreement (tracing quality). Agreement categories following standard thresholds: poor < 0.40, fair 0.40–0.59, moderate 0.60–0.74, good/excellent ≥ 0.75. Table 3Specific findings, including time to acquisition, quality scoring and parameter assessment of mobile devices compared to standard 12 lead ECG recordings. Correlation of Apple watch and KardiaMobile 6 L with standard 12 lead ECG using pearsons’ correlation coefficient.S12LApple WatchKardiaMobile 6 LICCAcquisition Characteristics Time to acquisition4 (2–10)15 (10–32)5 (4–10) Quality (3-point Likert scale)3 (2–3)2 (1–3)3 (2–3)Fleiss´κ 0.71Heart Rate & Cardiac Axis HRGold Standardr = 0.32, p < 0.001r = 0.44, p < 0.0010.67 QRS axisGold Standardr = 0.95, p < 0.001r = 0.96, p < 0.0010.93Intervals PR intervalGold Standardr = 0.51, p < 0.001r = 0.36, p < 0.0010.62 QRS durationGold Standardr = 0.22, p = 0.02r = 0.2, p = 0.130.58 QT intervalGold Standardr = 0.29, p < 0.001r = 0.64, p < 0.0010.81 QTcGold Standardr = 0.28, p < 0.001r = 0.47, p < 0.0010.55 R peak timeGold Standardr = 0.4, p < 0.001r = 0.53, p < 0.0010.6Voltages P wave voltageGold Standardr = 0.33, p < 0.001r = 0.49, p < 0.0010.57 R voltageGold Standardr = 0.74, p < 0.001r = 0.85, p < 0.0010.88 S voltageGold Standardr = 0.67, p < 0.001r = 0.76, p < 0.0010.84 T wave voltageGold Standardr = 0.65, p < 0.001r = 0.75, p < 0.0010.78Waveform Durations P wave durationGold Standardr = −0.04, p = 0.7r = 0.13, p = 0.330.41 T wave durationGold Standardr = 0.3, p < 0.001r = 0.26, p = 0.040.69Leg.: parameters for time to acquisition and tracing quality are given as (min.; median (range)). BMI = Body Mass Index, S12L = standard 12-lead ECG, PVCs = premature ventricular contractions, kg = kilogram, m = meters, r = Pearson correlation coefficient; values of time to acquisition and quality score are given as median (range); ICC = Intraclass Correlation Coefficient, two-way random-effects model, absolute agreement. Fleiss’ κ for categorical agreement (tracing quality). Agreement categories following standard thresholds: poor < 0.40, fair 0.40–0.59, moderate 0.60–0.74, good/excellent ≥ 0.75. Specific findings, including time to acquisition, quality scoring and parameter assessment of mobile devices compared to standard 12 lead ECG recordings. Correlation of Apple watch and KardiaMobile 6 L with standard 12 lead ECG using pearsons’ correlation coefficient. Leg.: parameters for time to acquisition and tracing quality are given as (min.; median (range)). BMI = Body Mass Index, S12L = standard 12-lead ECG, PVCs = premature ventricular contractions, kg = kilogram, m = meters, r = Pearson correlation coefficient; values of time to acquisition and quality score are given as median (range); ICC = Intraclass Correlation Coefficient, two-way random-effects model, absolute agreement. Fleiss’ κ for categorical agreement (tracing quality). Agreement categories following standard thresholds: poor < 0.40, fair 0.40–0.59, moderate 0.60–0.74, good/excellent ≥ 0.75. Acquisition time Significant differences were observed with regard to the time required to record ECG tracings with the AW, compared to S12L ECG, and KM. The median acquisition time for the AW (15 min, IQR 10–32 min) was notably slower compared to the other modalities (S12L ECG median 4 min, IQR 2–10 min; KM median 5 min, IQR 4–10 min), respectively. Tracing quality Qualitative assessment of ECG tracings revealed significant variations in tracing quality among the different devices. While the KM device demonstrated a median tracing quality score of 3 (range 2–3), the AW performed less well, showing a median quality score of 2 (range 1–3). The S12L ECG showed a median quality score of 3 (range 2–3). To complement median values, mean tracing quality scores (± SD) were calculated for all three modalities. The standard 12-lead ECG demonstrated a mean quality score of 2.8 ± 0.4, the KardiaMobile 6 L a mean score of 2.7 ± 0.5, and the Apple Watch a mean score of 2.1 ± 0.6. These results indicate modest but consistent differences in signal quality between devices, with the Apple Watch showing greater variability and a lower overall quality profile compared to the KardiaMobile 6 L and standard ECG. Quantitative parameters and correlation to S12L ECG Using Pearsons’ correlation coefficient, AW and KM tracings were compared to S12L ECG tracings. Very strong correlation (r value 0.80–1.00) was found for QRS axis (both AW and KM) and R voltage (KM). Strong correlation (r value 0.6–0.79) was found for R voltage (AW), S voltage (both AW and KM), T wave voltage (both AW and KM), and QT interval (KM). Moderate correlation (r value 0.4–0.59) was found for heart rate (KM), P wave voltage (KM), R peak time (both AW and KM), PR interval (AW), and QTc (KM). For the remainder of parameters, both devices showed low or very low correlation, with the lowest r values for P wave duration. See Table 3 for detailed results. Figure 2 provides Bland-Altman plots to illustrate correlations. Figure 3 provides example reconstructed 12-lead ECG from all three modalities in comparison (S12L, AW, KM). Fig. 2Bland-Altman plots of specific parameters with strong correlation. Bland-Altman plots of specific parameters with strong correlation. Fig. 3Example of 12-lead reconstructions from all three ECG modalities recorded from the same patient. Leg.: Reconstruction of all 12 standard ECG leads recorded with different devices in the same patient. ECG leads are labeled in standard nomenclature (I, II, III, aVR, aVL, aVF, V1-V6). KM Kardia Mobile 6 L, AW AppleWatch, S12L standard 12-lead ECG. Example of 12-lead reconstructions from all three ECG modalities recorded from the same patient. Leg.: Reconstruction of all 12 standard ECG leads recorded with different devices in the same patient. ECG leads are labeled in standard nomenclature (I, II, III, aVR, aVL, aVF, V1-V6). KM Kardia Mobile 6 L, AW AppleWatch, S12L standard 12-lead ECG. Arrhythmias In two participants we recorded solitary premature ventricular contractions (PVCs). Assessment of PVCs for origin suggested right ventricular outflow tract (RVOT) origin from S12L ECG, but also from AW and KM tracings in one patient. In the second patient such conclusions could not be made due to scarcity of PVCs. No other arrhythmias were recorded. Interobserver agreement Interobserver agreement was evaluated across the three blinded reviewers for tracing quality and quantitative ECG measurements. Agreement for tracing quality was good, with a Fleiss’ κ of 0.71 across all devices. Interobserver reliability for quantitative parameters showed similarly favorable results for robust ECG features. Intraclass correlation coefficients (ICC, two-way random effects, absolute agreement) were high for QRS axis (ICC = 0.93), R-wave voltage (ICC = 0.88), S-wave voltage (ICC = 0.84), and QT interval (ICC = 0.81). Parameters known to have lower signal amplitudes or higher artifact susceptibility in pediatric mobile ECG tracings demonstrated moderate agreement. ICC values were 0.62 for the PR interval, 0.58 for QRS duration, and 0.55 for QTc. As expected, agreement was lowest for P-wave duration (ICC = 0.41). Miscellaneous In two participants we were initially unable to record any tracing with the AW or KM devices. Both children participated in climbing exercise few hours prior to the appointment. The use of talcum powder had impaired electrodes to pick up signals from the skin. Only after thoroughly washing their hands twice, proper tracings could be acquired. Discussion This study demonstrates the technical feasibility of acquiring sequential 12-lead ECG tracings in pediatric patients using the Apple Watch and KardiaMobile 6 L. In summary, mobile ECG devices demonstrated good agreement with standard 12-lead ECG for selected robust parameters such as QRS axis, major ventricular voltages, and QT interval, while showing limited reliability for low-amplitude atrial signals and fine temporal measurements, highlighting both their feasibility and current diagnostic limitations in pediatric patients. These findings underscore that mobile devices cannot currently provide reliable assessment of all ECG domains in children, particularly those requiring precise temporal or low-signal measurements. Prolonged acquisition time with the Apple Watch One notable finding of this study was the significantly longer time required to acquire ECG tracings using the AW compared to the KM and S12L ECG. Several factors inherent to the design and operation of the AW likely contributed to this delay, reflecting both the challenges and limitations of its current application in pediatric cardiology. Sequential lead acquisition Unlike the KM and S12L ECG, which capture multiple leads simultaneously, the AW requires sequential acquisition of all 12 leads. This process inherently extends the time needed for a full 12-lead ECG recording. For each lead, the watch must be positioned at a specific location on the body and held steadily during the recording, which increases the logistical complexity and overall duration of the procedure. Strict recording requirements The AW demands a minimum of 30 s of high-quality signal recording for each lead before storing the ECG tracing. If poor signal quality is detected - due to patient movement, poor skin contact, or electrical noise signals - the device interrupts the recording process and restarts from the beginning. This sensitivity to signal quality, while aimed at ensuring reliable tracings, frequently prolongs acquisition time, particularly in pediatric patients who may struggle to remain still during the procedure12. Sensitivity to artifacts Tracing quality with the AW was observed to be inferior to that of both the KM and the S12L ECG. This difference is likely attributable to the AW’s heightened sensitivity to motion artifacts and environmental interference12,13. Pediatric patients, especially younger ones, are more prone to movement during ECG acquisition, which may lead to repeated interruptions and extended recording times. In contrast, the KM and S12L ECG systems appear more resilient to such artifacts, allowing for quicker and more consistent recordings. Heart rate variability and sequential acquisition effects Standard 12-lead ECGs and KardiaMobile 6 L record all leads simultaneously, producing a single physiologic HR measurement for that moment. In contrast, the Apple Watch requires sequential 30-second recordings for each lead, resulting in total acquisition times ranging from 10 to 32 min. During this period, children naturally exhibit fluctuating heart rates due to respiratory sinus arrhythmia, positional adjustments, and changes in cooperation or arousal14. As a result, the HR values extracted from individual Apple Watch leads represent different physiologic time points, which lowers correlation with devices that record all leads at once. Importantly, HR determined within each individual recording was accurate and interpretable across all modalities. The observed discrepancy therefore reflects physiologic variability rather than measurement error. However, prior studies have shown that during physical activity or tachyarrhythmias, heart rate measurements can vary significantly between devices15,16. Parameter-specific diagnostic limitations Parameters with low intrinsic signal amplitude (such as P-wave duration) or those highly sensitive to sampling variability (e.g., QRS duration) showed low or very low correlation. These findings highlight that not all aspects of pediatric ECGs can currently be assessed reliably using mobile devices. In particular, fine temporal measurements and low-voltage atrial signals appear more vulnerable to motion artifacts, sequential acquisition effects, and device-specific filtering. As a result, mobile ECG tracings should be used cautiously when evaluating subtle conduction abnormalities, atrial enlargement patterns, or QRS morphology refinements in children. Future pediatric-oriented hardware and software improvements will be necessary to expand diagnostic fidelity in these domains. Age-related heterogeneity An important consideration in the interpretation of our findings is the wide age range of the study cohort (4–17 years). Pediatric ECG parameters undergo substantial physiological and anatomical changes throughout childhood, including age-dependent variation in chest wall geometry, cardiac position within the thorax, normative ECG voltages, autonomic tone, and the magnitude of respiratory sinus arrhythmia. These factors introduce biological heterogeneity that may contribute to variability in tracing quality and diagnostic agreement. Although the study was not powered to perform formal age-stratified correlation analyses, qualitative inspection of the dataset suggested that younger children tended to exhibit more movement during acquisition, higher artifact burden, and greater heart-rate variability. These characteristics may in part explain the lower correlations observed for parameters with inherently high inter-individual variability, such as P-wave duration and QTc. Future studies with larger, age-balanced cohorts will be essential to more precisely quantify the influence of age on mobile ECG performance. Impact of cooperation on Apple Watch performance Cooperation level emerged as a particularly relevant factor for the Apple Watch, which requires sequential positional changes and sustained immobility during each 30-second recording. In younger children, limited cooperation resulted in repeated recording interruptions and prolonged acquisition times, increasing physiological heart-rate variability across sequential leads and contributing to noise and inconsistent signal quality. These factors likely influenced the device’s lower correlation coefficients for certain time-dependent parameters. This observation underscores the need for pediatric-specific acquisition workflows - or future device algorithms that are less susceptible to motion artifacts - to optimize smartwatch-based ECG recordings in younger age groups. Feasibility rather than diagnostic accuracy for arrhythmias Only two premature ventricular contractions were documented in the cohort, and while these could be visualized and morphologically assessed on both mobile devices as well as the standard ECG, this finding reflects technical feasibility rather than diagnostic accuracy. The present study was not designed or powered to evaluate arrhythmia detection performance, nor can conclusions be drawn regarding sensitivity or specificity. Larger studies with continuous or ambulatory monitoring would be required to meaningfully assess arrhythmia detection capability in pediatric populations. As expected, interobserver agreement was lower for parameters with inherently small signal amplitudes (e.g., P-wave duration) or greater sensitivity to artifact and sampling variability (e.g., PR interval, QRS duration, QTc). These observations align with the physiological challenges of pediatric ECG analysis and the technical limitations of sequential mobile ECG acquisition, and further underscore that certain fine-interval and atrial measurements remain less reliable when obtained from mobile devices. Clinical implications The prolonged acquisition time with the AW presents a potential barrier to its adoption in busy clinical settings, where efficiency is paramount. However, the sequential acquisition process and stringent recording requirements underscore the device’s focus on ensuring high-quality tracings. To optimize its utility, future iterations of the AW or accompanying software could benefit from enhanced signal processing algorithms to reduce sensitivity to motion artifacts, as well as streamlined workflows for multi-lead acquisition. In pediatric patients, strategies to mitigate these challenges could include engaging patients to remain still during recordings, using additional tools to stabilize the device, or tailoring the device’s algorithms to better handle pediatric-specific challenges such as smaller anatomical features and increased movement. Potential clinical applications of mobile ECG devices in arrhythmia diagnosis Our coincidental finding of premature ventricular contractions (PVCs) which could be successfully diagnosed originating from the RVOT from all three ECG modalities (AW, KM, S12L ECG) highlights the potential of mobile ECG devices not only for PVC diagnosis but also for broader applications in arrhythmia detection and management17,18. Conditions such as Wolff-Parkinson-White (WPW) syndrome, premature atrial contractions (PACs), and atrioventricular nodal reentrant tachycardia (AVNRT) could also benefit from mobile ECG technology. Future developments could expand their use by incorporating specific lead configurations tailored to diagnostic questions or electrical diseases. For example: ARVC Screening: Leads V1 to V3 could be emphasized for detecting epsilon waves or inverted T waves, both hallmarks of arrhythmogenic right ventricular cardiomyopathy (ARVC)19. QTc Assessment: Monitoring the QT interval in real-time could aid in diagnosing long QT syndrome20. Accessory Pathways: Mobile devices could be adapted to enhance the detection of delta waves in WPW syndrome21,22. ARVC Screening: Leads V1 to V3 could be emphasized for detecting epsilon waves or inverted T waves, both hallmarks of arrhythmogenic right ventricular cardiomyopathy (ARVC)19. ARVC Screening: Leads V1 to V3 could be emphasized for detecting epsilon waves or inverted T waves, both hallmarks of arrhythmogenic right ventricular cardiomyopathy (ARVC)19. QTc Assessment: Monitoring the QT interval in real-time could aid in diagnosing long QT syndrome20. QTc Assessment: Monitoring the QT interval in real-time could aid in diagnosing long QT syndrome20. Accessory Pathways: Mobile devices could be adapted to enhance the detection of delta waves in WPW syndrome21,22. Accessory Pathways: Mobile devices could be adapted to enhance the detection of delta waves in WPW syndrome21,22. Such adaptations would allow mobile ECG devices to function as targeted diagnostic tools, increasing their utility in both general cardiology and specific electrophysiological conditions. Currently, mobile ECG devices are particularly well-suited for diagnosing supraventricular tachycardia (SVT)23. Their portability and ease of use enable patients to capture tracings during symptomatic episodes without the need for immediate hospital visits. The rapid acquisition of ECGs facilitates timely rhythm identification, empowering patients to provide valuable diagnostic data to their physicians remotely. This is especially beneficial in pediatric and adolescent patients, who may experience infrequent or short-lived arrhythmias that would otherwise be challenging to document. Pediatric and adolescent acceptance The integration of mobile ECG devices with smartphone applications enhances their appeal in the pediatric and adolescent populations. This tech-savvy cohort is often highly receptive to wearable devices and health apps, fostering engagement and compliance24. By leveraging their familiarity with smartphones and wearables, these devices provide a non-invasive, user-friendly option for long-term monitoring and episodic rhythm assessment. Furthermore, the ability to share recordings with healthcare providers in real time reduces the logistical burden on families and promotes timely clinical intervention. Future directions To maximize their clinical potential, mobile ECG devices could benefit from: Improved Signal Processing: Enhancing the quality of tracings to reduce artifact sensitivity and improve diagnostic accuracy.Expanded Diagnostic Algorithms: Incorporating arrhythmia-specific features and tailored lead placements for targeted evaluations.Patient-Centered Design: Ensuring ease of use, particularly for younger patients or those with limited technical proficiency. Improved Signal Processing: Enhancing the quality of tracings to reduce artifact sensitivity and improve diagnostic accuracy. Improved Signal Processing: Enhancing the quality of tracings to reduce artifact sensitivity and improve diagnostic accuracy. Expanded Diagnostic Algorithms: Incorporating arrhythmia-specific features and tailored lead placements for targeted evaluations. Expanded Diagnostic Algorithms: Incorporating arrhythmia-specific features and tailored lead placements for targeted evaluations. Patient-Centered Design: Ensuring ease of use, particularly for younger patients or those with limited technical proficiency. Patient-Centered Design: Ensuring ease of use, particularly for younger patients or those with limited technical proficiency. Given the rapid development of artificial intelligence (AI) and machine learning, all of these future directions seem achievable within a foreseeable period of time. Efforts should be made to improve the diagnostic capabilities of mobile ECG devices not only for adults but also pediatric patients. Limitations This study was conducted at a single center and included a relatively small sample size, which limits generalizability. As a feasibility-oriented investigation, the findings should be interpreted as preliminary and hypothesis-generating rather than definitive assessments of diagnostic accuracy. The broad age range and modest cohort size also restrict our ability to evaluate device performance across pediatric subgroups. Furthermore, the broad age range of participants (4–17 years) introduces considerable developmental and physiological heterogeneity, which may affect ECG morphology, respiratory sinus arrhythmia, and cooperation during acquisition. These differences likely contributed to variability in tracing quality and agreement, particularly in younger children, who exhibited more movement and required repeated acquisition attempts with the Apple Watch. Larger, multicenter studies - including younger children and infants - will be necessary to confirm reproducibility, refine parameter-specific diagnostic accuracy, and better understand age-dependent variability in mobile ECG performance. Despite strong correlations for selected parameters, several essential ECG features (e.g. P-wave duration) showed weak agreement with S12L tracings. These limitations reflect both physiological factors (e.g. low signal amplitude of atrial activity in children) and device-related constraints. Although interobserver variability was assessed across three blinded reviewers, agreement varied by parameter, with lower reproducibility for low-amplitude atrial signals and fine temporal intervals. This highlights that, despite reasonable consistency for major diagnostic features, mobile ECG recordings - particularly those acquired sequentially - still present challenges for reliable assessment of subtle waveform characteristics in pediatric patients. Only two arrhythmic events were captured during the study, preventing any assessment of arrhythmia detection accuracy. Observations regarding arrhythmias therefore represent proof of feasibility rather than diagnostic validation. Consequently, mobile ECG devices cannot currently substitute for standard ECG when precise interval measurement or refined waveform analysis is required in pediatric patients. Finally, our findings apply to specific device generations and may not extend to future models. Conclusion This study highlights the growing potential of mobile ECG devices, such as the AW and KM, in pediatric cardiology. By evaluating their diagnostic accuracy and capabilities compared to the S12L ECG, we provide valuable insights into their strengths and limitations. While challenges remain, particularly with the AW, where sequential lead acquisition, sensitivity to artifacts, and strict recording requirements contribute to prolonged acquisition times and reduced tracing quality compared to the KM and S12L ECG. These limitations underscore the need for continued technological advancements to optimize signal quality, enhance diagnostic algorithms, and streamline workflows. Looking ahead, mobile ECG devices hold promise for expanding their role in targeted arrhythmia diagnostics in pediatric patients. By refining their functionality and tailoring their applications to specific clinical needs, these devices could significantly enhance diagnostic precision and patient care in pediatric cardiology.