The study was conducted as a 2-arm, randomized, observer-blinded intervention. The study’s acronym, TraceX, is derived from “TRansfer of AnatomiCal knowledge in the EXamination situation for preclinical medical students.” It was conducted as part of a curricular development project with the same name. The project was funded by the Medical Faculty of Tübingen University Hospital (Universitätsklinikum Tübingen [UKT]). The study protocol underwent an external peer-review process by reviewers not involved in the project. The dean of the faculty and the faculty commission approved the project in 2020.

Its protocol is illustrated in Figure 1, based on the SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) figure template [52]. The study compared students’ performance in a videotaped shoulder examination after a 3-week blended-learning course. The intervention and control groups received 2 different blended-learning modules.

The study took place at the Medical Faculty of Eberhard Karls University Tübingen in southern Germany. It was conducted between October 2021 and March 2022 by the Institute for General Practice and Interprofessional Health Care in cooperation with the Institute of Clinical Anatomy and Cell Analysis. In Tübingen, online courses are provided using the Integrated Learning, Information, and Work Cooperation System (ILIAS) online-learning platform, which is a commonly used learning management system among German universities (ILIAS open source e-Learning e.V). It allows log-in via authenticated student email accounts.

Several instruments used in the study were piloted in 2020 with 18 first-year medical students. The primary purpose of the pilot study was to check interrater reliability for the primary outcome measurement instrument and to calculate the internal reliability of the self-developed questionnaire items. The results of these checks and their impact on the development of the items are listed below for each respective instrument.

In addition, participants of the pilot study were asked to assess the online learning modules. Based on their feedback, the course content was adapted. Two of the pilot study participants provided valuable comments on course benefits and weaknesses in an additional voluntary online interview. The interview served as an extended evaluation. Furthermore, LG (who later conducted the group interviews in the main study) piloted interview guidelines and received interviewer training based on the participants’ feedback.

Recruitment for the main study started in May 2021. First-year medical students were approached via email and in 2 online lectures. Interested students were invited to contact the Institute for General Practice’s office to receive additional information (eg, type of study, goal, significance, and participant rights). Study participation was voluntary and did not impact other courses in the curriculum. For students who did not wish to participate in the study, regular curriculum activities proceeded. The teaching coordinator (responsible for course and lecture coordination at the Institute) communicated with students but did not participate in the study design or execution. Eligibility criteria included being a first-year medical student, having provided informed consent to use ILIAS, and feeling healthy enough to participate in the study.

Intervention and control courses were developed collaboratively in interdisciplinary workshops involving general practitioners (GPs), a medical didactics expert, an orthopedic surgeon, a psychotherapist, medical students, a physiotherapist, and a medical psychologist.

Common learning objectives were developed first and then applied to both groups, which differed in instructional design and learning format. The intervention group received access to structured online modules, including physiotherapist-led examination videos, functional and topographic anatomy videos performed by an anatomy lecturer, and modules on daily life impact and professional conduct. Content was integrated across preclinical and clinical domains. After self-study, students joined a 90-minute flipped-classroom seminar (taught by the same anatomy lecturer and a GP), which included preparation for a peer-training session. Due to COVID-19 restrictions, this 90-minute peer-training session was held in decentralized student pairs or moderated small groups on Zoom (Zoom Communications Inc), with trained peer tutors facilitating feedback.

The control group followed a more traditional format with lectures and self-study. Students received literature for self-study and instructions for self-examination (eg, palpation in front of a mirror). In addition, they attended 2 distinct synchronous online lectures: one on anatomy (anatomy lecturer) and one on clinical examination (GP), based on the same content as the intervention videos but presented with still images, self-demonstration by the lecturers, and a greater separation of clinical and preclinical content.

In summary, the intervention emphasized integrated, interactive learning with structured feedback and more practice time, providing a learning process with all 4 steps of Kolb’s learning cycle [15], while the control group reflected conventional, lecture-based learning with limited interaction and a more fragmented learning process. Please refer to Figure 2 for an overview.

From those students who signed a written informed consent to participate in the study, basic sociodemographic data were obtained at t0 to enable stratified randomization. Randomization was undertaken in a 1:1 ratio using lists generated with the R programming package (The R Core Team, R Foundation, and R Consortium). Randomization was stratified for gender and occupational experience and used block sizes of 2 and 4. Individualized identification codes were generated by the participants in the surveys (t1 and t4) to allow for linking of presurvey and postsurvey data without compromising participant anonymity or disclosing group affiliation. All other data that could reveal group affiliation were concealed until grading of the final examinations and statistical analysis was completed. Participants were not blinded to their own group affiliation. Randomization and group allocation were performed by 2 researchers not involved in the assessment and development of course content.

The groups received the same instructions for exam performance and grading criteria. Finally, a postcourse questionnaire was distributed to both groups, and voluntary postcourse online group interviews were conducted (t4). After study completion, members of the control group were provided full access to all e-learning resources to preclude disadvantage.

This study hypothesizes that an integrated course based on Kolb’s learning theory would improve first-year medical students’ performance in a standardized, complex OSCE shoulder station requiring structured clinical examination, anatomical knowledge, and professional conduct. Students in the intervention group attended the integrated course and were expected to outperform those in the control group. The primary outcome was OSCE performance at t3, assessed through a structured examination focusing on inspection, palpation, and functional movement assessment, with emphasis on identifying anatomical landmarks rather than detecting pathology.

Because contact restrictions prevented a face-to-face OSCE setting, participants recorded videos of their shoulder examinations and uploaded them to ILIAS. Video length was limited to 10 minutes and included a volunteer (peer, spouse, or family member) acting as a simulation patient. Students were instructed to ensure their volunteers had no preexisting shoulder pain or functional impairments and to obtain consent before recording. Simulation patients were asked to follow the students’ instructions and cooperate during the examination, creating a controlled environment intended to minimize bias introduced by the simulation patients.

The uploaded videos were assessed independently by a pair of blinded raters (GP and a medical student) using an examination checklist. The checklist comprised a total of 40 items, including 12 items on anatomical knowledge, such as topography and functional anatomy, 17 items on structured clinical examination skills, and 9 items on medical professionalism. It was developed in multidisciplinary workshops based on existing literature [14] and items from the OSCE used at UKT [53]. It was piloted in the pilot study. In that study, the pair of most congruent raters (a medical student and a GP) achieved an interrater validity of κ=0.573 (Cohen κ). Due to this moderate agreement, several items were operationalized for better standardization of examiners [54]. The final version of the examination score used in this study is included as Multimedia Appendix 1.

Participants’ subjective learning needs were assessed across three domains—anatomical knowledge, clinical examination skills, and professionalism—at time points t1 and t4 using a questionnaire. This questionnaire was developed during an interdisciplinary workshop and subsequently piloted in the initial study phase. The sample of 18 pilot study participants showed a Cronbach α of 0.92. The domain “anatomical knowledge” contained 5 items (ie, visible landmarks, anatomical structure of the shoulder, function, range of motion, and association of function with activities of daily living). The domain “clinical examination” also consisted of 5 items (ie, structured examination steps, history taking, shoulder inspection, shoulder palpation, and test of function). Finally, the domain “professionalism” was covered with 4 items (ie, use of patient-centered language, autonomous handling of the examination situation, perceptiveness toward patient feedback, and addressing patient feedback during the examination). All items were operationalized on a 4-point Likert scale. This yielded sum scores on a 5- to 20-point range and a 4- to 16-point range, respectively.

Theoretical anatomical knowledge was assessed before and after the course (at t1 and t4) using the same set of 10 randomized multiple-choice questions (MCQs). Each question had 5 answer options, with only 1 correct answer, resulting in a knowledge score ranging from 0 to 10. These MCQs reflect the current standard used in the written state medical examination in Germany.

Using Likert scale rating questions, we used 4 items of the standardized Affinity for Technology Interaction Short Scale (ATI-S), rendering a sum score on a 4- to 24-point scale. The ATI-S is a reliable (Cronbach α 0.88-0.92) and validated instrument [55].

The course evaluation included 5 items adapted from the standard evaluation used by the Medical Faculty of Tübingen, covering overall assessment, contribution to personal learning, curriculum alignment, course structure, and clarity of learning goals. Cronbach α for these items was 0.741 based on pilot study data. The evaluation was conducted online via ILIAS at time point t4. In addition, several items assessing blended learning formats were incorporated at t1 and t4, aiming to assess attitudes toward blended learning before and after the intervention. Students were asked to rate whether blended learning promotes self-directed learning, encourages lesson preparation and review, supports exam readiness, enhances curricular value, enables flexible learning, and increases study satisfaction.

Participants were additionally invited to online group interviews at t4. The interviews were analyzed with thematic analysis [56]. An in-depth mixed methods evaluation that includes the interviews will be published separately.

To evaluate the possible impact of the intervention on perceived stress, the validated and standardized 20-item Perceived Stress Questionnaire (PSQ-20) was applied at t4. It contains 20 items with 4 subscales: worries, tension, joy, and demands. Cronbach α is 0.80-0.86 [57].

The study was assessed by the Ethics Committee of the Medical Faculty of Tübingen University on May 4, 2020, (231/2020BO1) and did not require ethical approval. Study participation was voluntary and students provided informed consent to participate in the study and access the respective course material in ILIAS. No financial inducement was given for study participation. However, all participating students received a bookstore voucher to compensate for their time commitment. Furthermore, participants who completed the study requirements were eligible to enter a lottery drawing for one iPad. Data were stored on ISO 27001–certified servers at UKT. The trial protocol is available from the authors upon request.

Of 39 enrolled students, 38 were allocated to the intervention and control groups. In both groups, one participant, who, despite having signed informed consent, did not sign in to ILIAS. In the intervention group, one student dropped out before analysis due to an unresolvable video error. The control group had 4 dropouts who did not upload a video. In total, analyzable data from 31 study participants were obtained. Figure 3 below shows the study flow (CONSORT chart).

The majority of participants had no current employment. The intervention group had 2 more participants currently employed in the medical sector. Statistically, neither current nor past employment differed significantly between groups. Participant characteristics at baseline (t1) are presented in Table 1.

Both groups showed a significant reduction in self-reported learning needs in anatomy and clinical examination. MCQ performance improved significantly in both groups after the course, with no significant differences between groups. Participants in the intervention group spent, on average, 2 hours more in the learning module than those in the control group, a statistically significant difference. No relevant group differences were found in attitudes toward blended learning. In the course evaluation, the intervention group rated the blended learning experience and overall course evaluation less favorably but gave higher ratings for the achievement of learning goals and curricular alignment compared to the control group.

Table 3 provides an overview of the secondary outcomes. Group differences at t1 were not significant and are not included.

The 5 dropouts had higher subjective learning needs at t1, scored lower in the MCQ knowledge test, and had a slightly lower technical affinity. Their attitudes toward blended learning differed from those of the other study participants. While they believed that blended learning provided added value to learning, they were less confident that it helped them prepare for exams or increase personal satisfaction with their studies. The dropout sample was too small for statistical analyses. However, there are 2 ways to interpret this: there might be a subgroup of students with such low expectations of blended learning that they might not use the method if given the choice [60], or the small study population could reflect a selection bias of motivated, stress-resilient, technology-affine, high-performing students who would do well in any learning environment [32].

The intervention group required, on average, 2 additional hours to complete the course compared to the control group. This was due to engagement with video lectures and demonstrations, as well as participation in the peer-assisted synchronous learning session, all of which required learners to allocate additional time, even though they had greater flexibility to choose when and where to engage with most of the content than the control group. Similar increases in workload have been observed in other studies on blended learning formats [36].

This interpretation is further supported by the intervention group’s high ratings of course structure, clarity of learning objectives, and perceived alignment with the curriculum.

Although overall stress levels were high in both groups at t4, they were comparable to those of other medical student cohorts [20]. We therefore infer that the increased workload did not result in a heightened perception of stress among intervention group participants.

Technical affinity was equally high across both groups, consistent with findings that current medical student cohorts are generally digitally literate [27]. For our sample, this suggests that time invested in studying, rather than digital affinity, was more strongly associated with learning success [46,47].

This study is limited by its small sample size, although a power analysis indicated a need for 80 participants; only 38 were recruited, and 31 completed the trial. While randomization controlled for gender and professional experience, the small sample size limits generalizability and prevents definitive conclusions about causality.

Recruitment was difficult despite incentives, likely due to increased student workload during the COVID-19 pandemic [26,27]. Although results were promising, the recruitment shortfall and end of funding prevented a second cohort. In addition, curriculum changes during the later stages of the pandemic affected the study environment, leading us to conclude the trial and prioritize efforts to integrate the intervention into the curriculum.

Concerning the main outcome, the intervention course provided learners with tailored modular learning material and feedback opportunities, in contrast to the control group. Importantly, both groups received teaching on shoulder anatomy and clinical skills—a more integrated approach than the regular curriculum—and identical guidance on the performance assessment. This substantially reduces the study’s ability to isolate the effect of blended versus traditional learning. However, both groups showed knowledge gains, suggesting that early integration of clinical content is effective overall. The superior performance in the intervention group likely reflects not only the learning format but also the stronger alignment between the instructional design and the assessment tasks—consistent with our hypothesis that an integrated, theory-based course would improve clinical competency [47].

A reconnaissance effect on anatomical knowledge cannot be excluded, as the same MCQs were repeated at t4. However, student performance and learning needs indicated stable or improved knowledge. Peer learning sessions and exams were intended to be conducted face-to-face but were instead held via videoconferencing due to COVID-19. Although this deviated from the protocol, comparable learning outcomes have been reported in other studies. Video-recorded OSCEs are an established method [16], with only one upload failure. Interobserver reliability was substantial and consistent with other studies [9,14,37].

First-year medical students’ clinical performance on shoulder examinations can objectively be improved by an integrated blended-learning course in anatomy, professional conduct, and clinical examination skills. The results encourage early integration of clinical examination skills in preclinical medical education. Didactically, early clinical examination courses should focus on functional anatomy, a structured examination approach, and healthy subjects. Methodologically, the content can be delivered using blended learning that considers social aspects of learning and aligns the content to a standardized shoulder examination simulating a single OSCE station. If a face-to-face exam is not possible, videotaped examinations are a viable alternative. This study does not provide information about the sustainability of the learning. A cohort study that follows the progression through surface and deep learning phases to the transfer phase in this integrative approach is warranted [11].