This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mHealth and uHealth, is properly cited. The complete bibliographic information, a link to the original publication on https://mhealth.jmir.org/, as well as this copyright and license information must be included.
Atrial fibrillation (AF) screening using mobile single-lead electrocardiogram (ECG) devices has demonstrated variable sensitivity and specificity. However, limited data exists on the use of such devices in low-resource countries.
The goal of the research was to evaluate the utility of the KardiaMobile device’s (AliveCor Inc) automated algorithm for AF screening in a semirural Ethiopian population.
Analysis was performed on 30-second single-lead ECG tracings obtained using the KardiaMobile device from 1500 TEFF-AF (The Heart of Ethiopia: Focus on Atrial Fibrillation) study participants. We evaluated the performance of the KardiaMobile automated algorithm against cardiologists’ interpretations of 30-second single-lead ECG for AF screening.
A total of 1709 single-lead ECG tracings (including repeat tracing on 209 occasions) were analyzed from 1500 Ethiopians (63.53% [953/1500] male, mean age 35 [SD 13] years) who presented for AF screening. Initial successful rhythm decision (normal or possible AF) with one single-lead ECG tracing was lower with the KardiaMobile automated algorithm versus manual verification by cardiologists (1176/1500, 78.40%, vs 1455/1500, 97.00%;
The performance of the KardiaMobile automated algorithm was suboptimal when used for AF screening. However, the KardiaMobile single-lead ECG device remains an excellent AF screening tool with appropriate clinician input and repeat tracing.
Australian New Zealand Clinical Trials Registry ACTRN12619001107112; https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?id=378057&isReview=true
Consumer use of wearable technology capable of ambulatory assessment of heart rate and rhythm has significantly increased in recent years [
AF screening using single-lead ECG devices has been reported in hospital, primary care, and community settings with variable sensitivity and specificity [
The TEFF-AF study (registered with the Australian New Zealand Clinical Trials Registry [ACTRN12619001107112]) is an AF screening study conducted at the Soddo Christian Hospital (SCH). The SCH is located in the semirural town of Soddo in south-central Ethiopia, with a population of around 200,000 individuals. AF screening was undertaken by a team of 5 nursing and research support staff from the SCH following specialized training on the use of the KardiaMobile device, iPhone app (version 5.7.4, KardiaAI: 1.1.7), and online Research Electronic Data Capture database. The training included an initial tutoring session followed by subsequent hands-on practice in acquiring a best-quality single-lead ECG tracing with the KardiaMobile device. AF screening commenced at the SCH in August 2019 with inclusion criteria being any ambulant adult aged 18 years and above and able to provide informed consent. Signage in Amharic language was erected to advertise screening to aid recruitment (
Atrial fibrillation screening advertising (left) and study information (center, in Amharic language) and single-lead electrocardiogram recording (right).
All participants provided informed consent, and the study is approved by the SCH research ethics board. Baseline demographic and clinical parameters were obtained to characterize the cardiovascular risk profile of participating individuals. Measurements of height, weight, and blood pressure (Intellisense T5 automatic monitor, Omron Corporation) were obtained before single-lead ECG acquisition using the KardiaMobile device. As per the study protocol (
Atrial fibrillation (AF) screening protocol. ECG: electrocardiogram.
The KardiaMobile mobile single-lead ECG device records a bipolar lead I ECG tracing when 2 or 3 fingers from each hand of the user are placed in contact with the 2 electrodes (
The KardiaMobile ECG tracings obtained for the first consecutive 1500 participants in the TEFF-AF study were included in this analysis. Each single-lead ECG tracing has a rhythm determination by the KardiaMobile automated algorithm of normal, possible AF, bradycardia, tachycardia, unclassified, unreadable, or too short. Single-lead ECG traces were downloaded and analyzed independently by two cardiologists. The cardiologists also assessed diagnostic limitations for each tracing categorized as artefact, ectopy, bradycardia, tachycardia, or insufficient sample duration.
The dataset with deidentified information generated and analyzed during this study is available from the corresponding author on reasonable request.
Summary statistics were presented by frequency and percentage or mean and standard deviation as appropriate. Categorical data were analyzed using the chi-square test. Sensitivity and specificity for the ability of the KardiaMobile to produce a rhythm decision against the cardiologist ECG interpretation was calculated. Linear regression analysis was performed to assess the factors contributing to screening performance of the KardiaMobile automated algorithm. All statistics were performed in SPSS Statistics version 26 (IBM Corp), and statistical significance set at
A total of 1709 single-lead ECG tracings (including repeat tracing on 209 occasions) were analyzed from a cohort of 1500 participants who presented for AF screening. The baseline clinical parameters of the participants are shown in
Baseline clinical characteristics (n=1500).
Demographic and clinical information | Values | ||
Age in years, mean (SD) | 35 (13) | ||
Gender, male, n (%) | 960 (64.00) | ||
|
|||
|
Southern Nations, Nationalities, and Peoples’ Region | 1439 (95.93) | |
|
Omoria | 30 (2.00) | |
|
Amhara | 11 (0.73) | |
|
Other regions (including Somalia, B-Gumuz, Addis Ababa, Harar) | 19 (1.27) | |
|
|||
|
Orthodox | 416 (27.73) | |
|
Protestant | 988 (65.87) | |
|
Muslim | 70 (4.67) | |
|
Other or no religion | 22 (1.47) | |
|
|||
|
Illiterate | 55 (3.67) | |
|
Primary level school | 137 (9.13) | |
|
Secondary level school | 599 (39.93) | |
|
Certificate, diploma, or higher | 707 (47.13) | |
|
|||
|
Unemployed | 175 (11.67) | |
|
Employed | 682 (45.47) | |
|
Self-employed | 344 (22.93) | |
|
Others including student and retired | 297 (19.80) | |
|
|||
|
Height (cm) | 167.7 (8.6) | |
|
Weight (kg) | 67.1 (13.3) | |
|
Systolic blood pressure (mm Hg) | 124.0 (17.7) | |
|
Diastolic blood pressure (mm Hg) | 76.5 (11.7) | |
|
|||
|
Hypertension | 104 (6.93) | |
|
Diabetes mellitus | 34 (2.27) | |
|
Congestive cardiac failure | 20 (1.33) | |
|
Stroke | 3 (0.20) | |
|
Coronary artery disease | 2 (0.13) | |
|
Peripheral artery disease | 0 (0.00) | |
|
Chronic lung disease | 16 (1.07) | |
|
Chronic renal disease | 5 (0.33) | |
|
Valvular heart disease | 11 (0.73) | |
|
Obstructive sleep apnea | 2 (0.13) | |
|
Thyroid disease | 21 (1.40) | |
|
Smoker | 5 (0.33) | |
|
Khat/alcohol use | 14 (0.93) | |
|
Infectious disease | 288 (19.20) |
Of the initial single-lead ECG tracings from 1500 participants, the KardiaMobile algorithm was unable to provide a rhythm decision in 21.60% (324/1500) due to unclassified (130/1500, 8.67%), tachycardia (128/1500, 8.53%), unreadable (62/1500, 4.13%), too short (3/1500, 0.20%), and bradycardia (1/1500, 0.07%). Representative examples of these tracings are shown in
Examples of KardiaMobile single-lead electrocardiogram tracings.
The KardiaMobile automated algorithm successfully obtained a rhythm decision on the first attempt for 78.40% (1176/1500) of participants, which was considerably lower than manual assessment by cardiologists (1455/1500, 97.00%;
Comparison of KardiaMobile algorithm versus manual assessment by cardiologists. ECG: electorcardiogram.
KardiaMobile automated algorithm versus cardiologists’ adjudication for single-lead electrocardiogram (ECG) for rhythm decision in n=1500 participants and atrial fibrillation detection in n=1709 ECG tracings.
KardiaMobile algorithm | Cardiologists’ adjudication | ||||||||
|
Rhythm decision | Atrial fibrillation | |||||||
|
Yes | No | Yes | No | |||||
|
|
|
|
|
|||||
|
Yes | 1168 | 8 | —b | — | ||||
|
No | 287 | 37 | — | — | ||||
|
|
|
|
|
|||||
|
Yes | — | — | 9 | 61 | ||||
|
No | — | — | 3 | 1636 |
a
bNot applicable.
c
In total, 154 participants met criteria for a 12-lead ECG, but this was obtained in only 59.09% (91/154) due to participants not wanting to wait for the 12-lead ECG to be performed in the SCH emergency room. However, upon review of the single-lead ECGs meeting study criteria for a 12-lead ECG to be performed, the cardiologists adjudicated 89.61% (138/154) of these single lead ECGs to be of adequate quality for a rhythm decision. In total, diagnoses from the 12-lead ECGs were 89.01% (81/91) sinus rhythm, 1.10% (1/91) supraventricular tachycardia, and 9.89% (9/91) AF.
We analyzed the performance of the KardiaMobile automated algorithm for providing an initial rhythm decision. There was a linear relationship between ongoing participant recruitment and the occurrence of a no rhythm decision from the initial KardiaMobile tracing (
Cumulative occurrence and contributors to no rhythm decision from KardiaMobile’s automated algorithm on initial electrocardiogram tracing: (A) cumulative occurrence of no rhythm decision from initial electrocardiogram tracing and (B) occurrence of unreadable tracing was significantly reduced when compared with unclassified and tachycardia tracings with increasing patient recruitment.
This study evaluated the utility of the KardiaMobile single-lead ECG device for AF screening in a semirural Ethiopian population of 1500 individuals from the TEFF-AF study. We found the KardiaMobile device performance to be suboptimal with successful automated rhythm decision following a single ECG trace of only 78%. This yield increased to 87% following a second KardiaMobile ECG tracing. As experience increased with ongoing patient recruitment, we encountered significant reduction in unreadable tracings. The ongoing occurrence of tachycardia and unclassified tracings contributed largely to the automated KardiaMobile algorithm’s inability to achieve successful rhythm decision. In contrast, manual cardiologist assessment was able to obtain a rhythm decision in almost all cases (97%) with a single ECG. Taken together, our findings suggest that manual physician input remains necessary when the KardiaMobile device is used for AF screening.
The use of single-lead ECG devices is of increasing interest given the potential benefits of portability and scalability. Furthermore, automated rhythm analysis may allow for the use of such devices by individuals without formal medical training. However, there are limited data on the accuracy of these devices and their automated rhythm analysis algorithms in such settings despite the KardiaMobile device having been FDA-approved since 2012. In a small validation study, the KardiaMobile’s automated AF detection algorithm was reported to yield high sensitivity of 98% and specificity of 97% with overall accuracy of 97% [
Recently, several studies have reported on the use of other smart wearable devices using photoplethysmography-based technology for AF screening. The Apple Heart Study reported on the ability of a smartwatch photoplethysmography sensor and algorithm to screen individuals for an irregular pulse. Of 419,297 individuals, 2161 (0.52%) had a smartwatch-detected irregular pulse, with AF confirmed in 34% of those who returned an ECG patch. Of the individuals who had a smartwatch-detected irregular pulse while simultaneously wearing an ECG patch, 84% (78/86) were in AF at the time [
Our study has important clinical implications for AF screening and highlights opportunities for future research. Prior research has shown that automated device algorithms can achieve accurate rhythm analysis under ideal conditions. However, our real-world experience in a resource-limited setting demonstrates that single-lead ECG tracing artefact and other limiting factors frequently prohibits algorithm interpretation. Despite limitations with tracing quality, manual cardiologist adjudication can still provide a rhythm diagnosis in the vast majority of cases. Thus, our findings suggest that physician input remains necessary for AF screening until further improvements in automated algorithms occur. In the meantime, repeat ECG tracings and increasing familiarity with using single-lead ECG devices are helpful to reduce unreadable tracings to improve diagnostic yield. Future studies should be undertaken to validate other mobile device technology and automated algorithms in real-world settings.
Our screening protocol required a repeat tracing for occasions without a rhythm decision. However, this was not performed in a proportion of the participants, resulting in an incomplete data set. We acknowledge that the clinical value of AF screening in a young cohort with unknown risk factors for stroke is unclear. Nevertheless, given the knowns and unknowns of AF in sub-Saharan Africa and the higher prevalence of rheumatic heart disease, we did not restrict the AF screening to the typical target population of older individuals with higher stroke risk in developed countries [
The performance of the automated algorithm of the KardiaMobile single-lead ECG device was suboptimal when used for AF screening. However, the KardiaMobile device remains an excellent and affordable tool when used in low-resource settings with appropriate clinician input.
atrial fibrillation
electrocardiogram
US Food and Drug Administration
KardiaMobile
Soddo Christian Hospital
The Heart of Ethiopia: Focus on Atrial Fibrillation study
BP is supported by a Doctoral Scholarship from the Hospital Research Foundation. CW is supported by a Postdoctoral Fellowship from the National Heart Foundation of Australia. CW and DL are supported by Mid-Career Fellowships from the Hospital Research Foundation. PS is supported by a Practitioner Fellowship from the National Health and Medical Research Council of Australia and by the National Heart Foundation of Australia.
All authors have full access to all the data and take full responsibility for the integrity of the data and the accuracy of data analysis. BP, SHC, AC, and DHL were responsible for study design and conception. BP, SHC, CXW, AJ, SI, GT, and AC were responsible for data acquisition and analysis. BP, CXW, AJ, SI, GT, PS, and DHL interpreted the data. BP, CXW, PS, and DHL drafted and revised the manuscript.
CXW reports that the University of Adelaide has received on his behalf lecture, travel, and/or research funding from Abbott Medical, Bayer, Boehringer Ingelheim, Medtronic, Novartis, Servier, and St Jude Medical. PS reports having served on the advisory board of Medtronic, Abbott Medical, Boston Scientific, Pacemate, and CathRx. PS reports that the University of Adelaide has received on his behalf lecture and/or consulting fees from Medtronic, Abbott Medical, and Boston Scientific. PS reports that the University of Adelaide has received on his behalf research funding from Medtronic, Abbott Medical, Boston Scientific, and MicroPort CRM. DHL reports that the University of Adelaide has received on his behalf lecture and/or consulting fees from Abbott Medical, Bayer, Boehringer Ingelheim, Biotronik, Medtronic, MicroPort CRM, and Pfizer.