This was a prospective, self-controlled, single-center study that recruited doctors and medical students interested in free cardiac auscultation training through multiple channels within our hospital and affiliated medical school. The recruitment campaign was coordinated by the hospital’s Department of Medical Education and disseminated via several official WeChat groups, including those for undergraduate medical students, staff physicians, and visiting trainees. In addition, printed recruitment posters containing a QR code for registration were displayed in various teaching and clinical areas of the hospital and medical school.

Interested individuals could scan the QR code to access an online registration form. The form collected basic demographic information, participants’ prior experience with cardiac auscultation, and their motivation for enrolling in the course. Enrollment was open until the maximum capacity of the online classroom was reached, resulting in a total of 199 registrants on a first-come, first-served basis.

This study was approved by the Institutional Review Board of Peking Union Medical College Hospital (I-23PJ1679). During registration, all prospective participants were presented with an electronic informed consent form that described the study purpose and procedures, explained that the online teaching sessions would be recorded, and noted that anonymized data might be used for research analysis. Only those who read and electronically signed the consent form were able to complete registration and participate in the course. Participation was entirely voluntary, and no financial or material compensation was provided.

Continuous variables were assessed for normality using the Shapiro-Wilk test. Normally distributed data were presented as means (SDs), while nonnormally distributed data were expressed as medians (IQRs) or median (minimum, maximum). Categorical variables were presented as frequencies and percentages. Nonnormally distributed data were analyzed using the Wilcoxon signed rank test. Differences in categorical variables were analyzed using either the χ² test or Fisher exact test, as appropriate.

Spearman correlation coefficients and multivariate linear regression analyses were used to identify factors influencing final quiz scores. Variables with Spearman correlation coefficients (P<.05) were included in the multivariate regression model. Collinearity diagnostics were performed, and variables inducing collinearity were excluded. Factors associated with excellent learners were evaluated using receiver operating characteristic (ROC) curve analysis.

All statistical analyses were performed using SPSS Statistics for Windows, Version 23 (IBM Corp) and GraphPad Prism, Version 10.1.2 (GraphPad Software). Statistical significance was defined as P<.05.

A total of 199 individuals voluntarily registered for the training, including 122 doctors and 77 medical students. The recruitment of doctors was completed within 2 days, whereas student recruitment required about a month. All registrants indicated that they had not yet mastered the skill of cardiac auscultation. The doctors were significantly older than the students and had substantially more years of prior exposure to auscultation learning. From registration to actual participation, a higher proportion of doctors attended the training compared with medical students (96/122, 78.7% vs 50/77, 64.9%; χ²₁=4.57; P=.03). Overall, 73.4% (146/199) of registrants participated in the training, corresponding to a 26.6% (53/199) dropout rate. However, only 23.1% (46/199) of all registrants attended all four sessions (Figure 1), with no significant difference in full participation between doctors and students.

For participation motivation, doctors were more driven by intrinsic motivation than medical students (89/122, 73.0% vs 15/77, 19.5%; χ²₁=54.10; P<.001). Intrinsic motivation was significantly correlated with age (r=0.394; P=.004) and total study time (r=0.145; P=.04), but not with the number of in-class random questions answered (r=0.153; P=.11). A χ² test further indicated that intrinsic motivation was associated with full participation (χ²₁=4.03; P=.045). A significant increase in posttraining scores suggested that the program was effective (Table 1).

During the four sessions, a total of 16 random questions were posed, and 11 sets of review materials were distributed online after each class. The system automatically recorded the number of in-class responses and the frequency of postclass access to review materials. Of the 146 training participants, 49 (33.6%) participants watched the lecture videos after class; however, only 10 (6.8%) viewed all four full recordings. The lecture material was reviewed by 111 (76.0%) participants, but only 20 (13.7%) accessed all available materials.

The duration of postclass video viewing varied substantially among participants. Age was significantly and positively correlated with total study time (r=0.366; P<.001), time spent reviewing recorded videos (r=0.330; P<.001), and the number of review materials accessed (r=0.355; P<.001; Table 2).

We performed univariate correlation analyses to examine the relationships between multiple factors—including learning motivation, overall participation, number of class attendances, duration of class participation, frequency of in-class responses to random questions, number of postclass material reviews, number of times watching lecture videos, and total study time—and the final scores. Significant correlations were observed between the final scores and full participation (r=0.351; P=.02), frequency of in-class responses (r=0.431; P=.004), and number of postclass material reviews (r=0.345; P=.03).

Age, participant type (doctor or student), years of prior auscultation study, motivation for participation, and time spent attending live sessions were not significantly correlated with final scores. Further multivariable linear regression analysis showed that only full participation and frequency of in-class responses remained independently associated with final scores, whereas postclass material review frequency did not significantly affect performance (Table 3).

Six participants increased their scores by more than 60 points after the training, placing them in the top 10% of all participants and defining them as excellent learners. ROC curve analysis suggests that total study time, actively answering questions in class, full participation, and the extent of postclass review were all significantly related to achieving excellent learner status (Figure 2).

In this voluntary noncredit online cardiac auscultation course, cohort-level test scores improved from pre- to posttraining, indicating overall instructional effectiveness. Nevertheless, only 23% of registrants completed all four sessions, underscoring the difficulty of sustaining participation in voluntary settings. Within the cohort, intrinsic motivation was positively associated with full participation, and both full participation and more frequent content‑aligned in‑class responses were independently associated with higher final scores. Postclass review was not an independent predictor in multivariable models, yet it consistently characterized learners who achieved excellent performance in ROC analyses.

Taken together, the data suggest a plausible pathway for success in voluntary online learning: autonomous motivation→sustained participation→active content‑aligned engagement→higher performance. This pathway is consistent with the self‑determination theory and with contemporary applications of autonomy‑supportive design in health professional education, which emphasize persistence and self‑regulation in learning tasks [20,21]. Methodologically, our study contributes by operationalizing autonomy through objective behavioral traces—attendance, periodic in‑class responses, and review—rather than relying solely on self‑report, which aligns with recent work mapping measurable dimensions of student engagement in health professional education [22].

Low completion in our cohort mirrors patterns reported in large‑scale voluntary online courses, where completion is often modest; for example, completion of massive open online courses frequently clusters around 10% in some contexts. A recent systematic review concluded that dropout is multifactorial and more closely related to course attributes, learner behaviors, and motivation/self‑regulation than to content alone, reinforcing the need for proactive planning to sustain participation in voluntary courses [11]. In practical terms, clear expectation setting before enrollment—regarding time commitment, learning objectives, and assessment schedule—and simple scheduling reminders have been reported to improve learner persistence in online education [11] and may help participants maintain full engagement in voluntary courses.

Beyond attendance, maintaining focused attention during sessions appears salient. Empirical work shows that off‑task device use is more frequent online than face‑to‑face and is associated with perceived distraction, which can erode engagement [23]. In medical education contexts, survey evidence and expert opinion further suggest that attention tends to decline after approximately 10-15 minutes unless refreshed through interactive elements. Accordingly, periodic content-aligned checks (brief questions or polls), as implemented in our study, are advisable in synchronous sessions [24].

Finally, postclass review may play a selective role in mastery. Although postclass review was not an independent predictor in our adjusted model, high performers engaged in postclass review more often. This pattern is consonant with meta-analytic evidence that spaced or retrieval-based reviews improve long-term retention and, in some cases, clinical behavior change relative to massed learning [25]. In health professional education, spaced e‑learning has been shown to enhance knowledge retention in basic life support compared with single‑session learning [26]. Future implementations could explore embedding simple, low-burden review structures—such as a 48-hour recap and a 1-week revisit—and providing automated reminders to prompt learners to consolidate key material.

Notably in our cohort, the associations between participation/engagement and performance were not explained by measured demographic characteristics (age, years of exposure to auscultation, workplace), suggesting that the behavioral pathway outlined above could be applicable across subgroups in similar settings.

This study has several limitations. It was conducted at a single center with voluntary participation, which may limit generalizability and introduce self-selection bias. The primary outcome was a short-term test score, and long-term retention was not assessed. Although the pre- and posttests followed the same blueprint, using different item sets may have introduced measurement variability. Motivation was assessed by a single unvalidated item, and the voluntary single‑center sample entails self‑selection; hence, residual confounding and limited generalizability remain.

Learners who consistently attended all sessions, stayed focused during class, and reviewed material after class achieved the best results in this voluntary online clinical skills course. These behaviors represent practical expressions of learner autonomy and can be strengthened through simple habits—committing to scheduled sessions, engaging actively with class tasks, and revisiting key materials on a spaced schedule. For busy health professionals, such disciplined learning routines may transform limited study time into durable competence.

Beyond this specific course, these findings highlight the importance of quantifying and supporting learner autonomy as a measurable construct in online medical education. Incorporating behavioral analytics to monitor participation, engagement, and review patterns may provide educators with actionable insights to design more autonomy-supportive learning environments. Future research should investigate how these learner-driven behaviors can be further supported through course design and motivation-enhancing strategies to sustain long-term engagement and performance across diverse educational settings.