Study participants underwent a screening examination procedure, as shown in Figure 1. From each study participant, demographic data and ECG, PCG, and echocardiographic data were collected according to International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use good clinical practice. This is a widely adopted internationally agreed upon standard for designing, conducting, recording, and reporting clinical trials involving human participants. No clinical evaluation, including blood tests or swab samples related to ARF, were performed. A total of 1385 participants gave consent participants, of whom only 584 (42.2%) completed the ECG, PCG, and echocardiographic screening procedures, whereas the remainder missed one or more of the examinations mainly due to conflicting schedules, anxiety about echocardiographic screening, peer pressure, or unspecified personal reasons.

ECG data were recorded using a commercial CE-marked device, KardiaMobile 6L (AliveCore, Inc). KardiaMobile is small pocket device that is connected via Bluetooth to a compatible smartphone that has the Kardia app from AliveCor, Inc, installed, shown in Figure 2A. The device has 3 electrodes with a sampling frequency of 300 Hz. For obtaining an ECG recording, participants must touch 3 electrodes on the device, 1 on the back with their left leg and 2 on the front with the fingers of the left and right hand. ECG recordings were obtained at least twice, both from the knee or ankle positions, and repeated if deemed necessary (Figure 3C-D). Each recording had a duration of 30 seconds.

The temporal evaluations on ECG waveforms were performed manually by 2 physicians. After performing a 50-Hz powerline interference and Butterworth zero-phase band-pass filter from 1 to 100 Hz, an averaged ECG was computed by aligning and averaging the cardiac cycles in the lead II waveform using the R wave as a reference point. Three ECG intervals, PR, QRSd, and QTc, were then evaluated for RHD-positive participants from an averaged ECG. Prolonged changes in these intervals have been reported in patients with RHD [24-26]. The intervals were normalized using the heart rate (HR) and calculated considering a threshold margin of –10 to +10 ms for the QRSd interval and –15 to +15 ms for the PR and QTc intervals. HR-adjusted PR intervals [29] were used to determine PR prolongation. QTc values for each heartbeat were calculated using the formula by Bazett [30]. The QRS morphology was classified as normal if it was within the range of less than 105 ms considering the age of the cohort [31].

PCG data were recorded using a Thinklabs One stethoscope, a CE-marked medical device for auscultation. The stethoscope was connected to a PC, and input cardiac sounds were recorded using the Audacity open-source software (version 3.7.2; Muse Group) [32], as illustrated in Figure 2B. A wideband (20-2000 Hz) device mode audio filtering was set to acquire the heart sounds sampled at 44.1 kHz. The recording was taken from 4 positions: aortic, pulmonic, tricuspid, and mitral areas, as shown in Figure 3B.

The PCG recordings of mild regurgitations from participants with a BMI above 24 kg/m² were not audible, so different positioning strategies, such as left lateral decubitus, sit-up-lean-forward, and supine, were used to obtain a better signal. A total of 60 seconds of cardiac sound data were recorded from the aortic, pulmonic, tricuspid, and mitral areas. The recorded auscultations were played back and listened to by 2 physicians. Any pathological murmur during systole over the apex was considered an MR. Diastolic murmurs were labeled as aortic regurgitation (AR) or MS based on the intensity and duration of the murmur. Any other murmur was labeled as “other.”

A comprehensive transthoracic echocardiogram was performed by trained cardiologists using a portable Vivid iq (GE Vingmed Ultrasound) equipped with an M5Sc 2.2-MHz transducer (Figure 3C). Image acquisition was performed with the patient in a left decubitus position. Cardiac size and function, as well as the site and severity of the valvular lesion, were assessed using B-mode, spectral Doppler, and color Doppler imaging. We recorded clips with 3 to 5 cardiac cycles or 10 cycles in the case of atrial fibrillation. Afterward, the recorded images were independently reviewed by 2 cardiologists, and decisions were made using the WHF 2012 criteria.

The WHF 2012 criteria [2] (Multimedia Appendix 1) detail the diagnosis for definite RHD, borderline RHD, and normal echocardiographic findings in A, B, C, and D evidence groupings. Definite RHD requires either pathological regurgitation (MR or AR) combined with at least 2 morphological features of RHD in the affected valve, MS with a mean gradient of 4 mm Hg or higher, or concurrent borderline features in both the MV and aortic valve (AV). Borderline RHD is characterized by isolated morphological valve abnormalities or isolated pathological MR or AR without meeting the full criteria for definite RHD. Borderline RHD diagnosis cannot be based on a single abnormality; it requires either mild pathological regurgitation accompanied by at least one morphological feature or multiple morphological features of valvular involvement. Normal findings may include physiological regurgitation and isolated morphological changes without associated pathological stenosis or regurgitation. Participants with other nonrheumatic findings or comorbidities were excluded from the final analysis. Participants and/or their legal representatives were notified about findings indicative of RHD. Furthermore, all RHD-positive patients were referred to a nearby hospital for further guideline-recommended clinical management.

This study investigated the current prevalence of asymptomatic RHD among students aged 10 to 20 years in underserved areas in Ethiopia. The potential use of affordable clinical examination for periodic surveillance via PCG and ECG was assessed on echocardiography-confirmed cases. Our findings showed that there is a persistent high prevalence of RHD in at-risk communities, with female predominance. Echocardiography revealed MV involvement as the most common manifestation of RHD, typically presenting with leaflet thickening and restricted tip motion. Definite participants exhibited characteristic murmurs, whereas borderline or asymptomatic participants often showed nonspecific auscultatory findings. Low-cost ECG and PCG were insufficient for early detection as PR interval prolongation did not reliably indicate early-stage disease, although this prolongation was more prominent in advanced chronic participants [24-26]. These results underscore the importance of echocardiography-based screening for timely identification and management of subclinical RHD.

RHD is highly prevalent among students in endemic regions, affecting female individuals more often [1,2]. This study also found an asymptomatic RHD prevalence rate of 32.5 per 1000 population among students, which is consistent with reports from Ethiopia (Multimedia Appendix 1). However, our figures are slightly higher than the pooled prevalence figures from other similar studies in LMICs [10,12]. The findings provide further evidence of a nondecreasing trend in RHD burden in high-risk areas, with persistent transmission of group A Streptococcus infections and gaps in primary prevention.

As can be observed in Table 2, the RHD burden was 60% higher in female than in male individuals. Most of the RHD-positive participants (16/19, 84.2%) were older than 15 years, which aligns with trends observed in previous studies [10]. Overall, the asymptomatic presence of early-stage RHD in school-aged children and adolescents was not strongly correlated with household size, age, or sex in this cohort. While elevated prevalence in female individuals has historically been linked to childbearing age susceptibility [21], our data suggest elevated rates among even younger school-aged girls, which prompts further investigation into the potential role of biological, environmental, or sociocultural risk factors. Disease severity also correlated with advancing age, likely reflecting cumulative valvular damage from untreated ARF recurrences, as described in longitudinal cohorts [11]. However, the pathophysiological basis for sex disparities remains unclear. Importantly, while overcrowding was reported to have a positive and strong association with the prevalence of ARF and subsequently RHD, no strong associations were found for socioeconomic factors such as household income, parental occupation, or literacy levels [15]. This reinforces the necessity of population-wide screening irrespective of demographic proxies in disease-prone regions for early diagnosis and timely treatment of group A Streptococcus infections.

Valvular involvement patterns in our cohort align with global epidemiological profiles, where the MV is most frequently affected, followed by the AV, consistent with reports from Australia, New Zealand, and Uganda [19,20]. This sequence of valve involvement is likely attributed to the MV’s hemodynamic vulnerability during ARF, where inflammatory insults preferentially target valve leaflets and chordae [17]. Leaflet thickening and restricted motion dominated MV pathology, confirming the revised 2023 WHF RHD diagnostic criteria. As reported in the study by Webb et al [19], isolated MR was more prevalent than MS in asymptomatic participants, thereby supporting the paradigm of early-stage rheumatic damage characterized by inflammatory retraction and chordal elongation rather than fibrotic fusion. However, combined MR and AR occurred in 84.2% (16/19) of the participants in this study, which is consistent with previously reported findings [11]. Conversely, isolated AR was uncommon in this study cohort, with a single case reported. This is inconsistent with the results reported in the study by Rothenbühler et al [11], who found AR prevalence to be 21%. This discrepancy may reflect regional variations in ARF severity, delayed diagnosis, or genetic susceptibility to multivalvular involvement [4-6].

When these patterns of structural and functional involvement are interpreted within the framework of echocardiographic staging of rheumatic valvular disease in the 2012 WHF criteria [2], early inflammatory valvular changes can be distinguished from chronic advanced RHD. Borderline RHD is suggested by isolated rheumatic morphological abnormalities without pathological regurgitation, such as mitral leaflet thickening, posterior mitral annular thickening, restricted leaflet motion, or irregular or focal thickening of the AV. Definite RHD is suggested when these characteristic morphological changes are accompanied by pathological MR and/or AR, indicating established RHD with functional hemodynamic consequences. However, prolonged PR intervals were only observed in participants with borderline RHD (2/19, 10.5%), not in those with definite RHD.

It is noteworthy that PCG screening detected pathogenic audible murmurs in approximately two-thirds of the participants (12/19, 63.2%), which surpassed the levels previously reported in auscultation-based studies [28]. This supports the hypothesis that subclinical RHD generates subtle murmurs associated with early valvular changes without a clinically pathological murmur [2]. However, in endemic regions, clinically silent RHD is 7 to 8 times more prevalent than symptomatic RHD [11]. This highlights the limitations of murmur-dependent screening and underscores the need for multimodal data evaluation and echocardiography as the gold standard.

The ECG findings revealed QRSd and QTc interval prolongation in RHD-positive participants, consistent with conduction abnormalities attributed to myocardial fibrosis or inflammatory blockages [24-26]. These changes exhibited a correlation with the timing of ARF and disease severity. However, these results differed from those of studies that emphasized arrhythmias due to atrial ectopy and heart block in advanced RHD [24,26]. Atrial ectopy increases the risk of atrial fibrillation, whereas heart block typically produces bradyarrhythmia due to impaired electrical conduction. This is because rhythmic disturbances were minimal in our early-stage cohort. This observation suggests that arrhythmic manifestations may emerge at a later stage in the disease progression, although longitudinal data are necessary to substantiate this claim.

In summary, our findings advocate for integrated, technology-enhanced screening frameworks to combat RHD in resource-limited settings. Recent advancements in wearable ultrasounds for cardiac imaging have demonstrated potential to make imaging more cost-effective while maintaining comparable performance to that of standard portable echocardiography devices [33]. Hu et al [33] proposed an automated approach that uses deep learning algorithms to estimate the volume of the left ventricle and interpret B-mode ultrasound images with an intersection over union score above 85%, thereby facilitating more precise and efficient analysis.

Furthermore, the evaluation of MR jet length as a single criterion to detect RHD has shown sensitivity between 73% and 78.9% and specificity of 82.4% to 87.3% [4]. The application of such wearable devices with simplified criteria in combination with low-cost PCG or ECG sensors could improve diagnosis and reduce global RHD-associated burden. The development of such a cost-effective and scalable approach relies on the availability of large, diverse, and multimodal datasets capable of capturing the complex and heterogeneous features of RHD to support accurate and timely diagnosis. Future studies should explore using predictive modeling of multimodal ECG, PCG, and echocardiogram data to stratify the risk of RHD progression and evaluate cost-effective interventions in real-world settings.