The report of this study followed the recommendations of the Simulation-Based Research Extensions for the CONSORT (Consolidated Standards of Reporting Trials; Checklist 1) Statement [20] and the Modified Medical Education Research Study Quality Instrument [21].

This was a parallel-group randomized controlled trial with a pretest-posttest design comparing collaborative and individual simulation-based learning over 3 weeks in undergraduate physical therapy students (Figure 1). Participants were randomly assigned (1:1) to either collaborative or individual simulation-based learning and completed standardized patient assessments before and after the intervention. The learning assessment consisted of an individual standardized patient encounter, from which the primary outcome (clinical competence) and secondary outcomes (communication and therapeutic alliance) were evaluated. The study was conducted in the rehabilitation laboratories of the Faculty of Health Sciences of the Catholic University of Temuco, Chile.

Nonprobabilistic convenience sampling was used for this study. All eighth-semester physical therapy students (N=48) enrolled in the Musculoskeletal Rehabilitation course during 2022 were screened for eligibility. The inclusion criterion required the successful completion of the course. An independent researcher, who held no grading responsibilities or administrative ties to the students, conducted an informational session in December 2022. During this meeting, the study objectives and simulation-based experiences were detailed, emphasizing that participation was strictly voluntary and would result in no academic prejudice. Of the 48 eligible students, 31 (64.5%) provided written informed consent and were enrolled. Data collection and clinical simulation training took place between December 2022 and January 2023, coinciding with the start of the summer break.

Participants were randomly assigned to either an experimental group with collaborative simulation-based learning or a control group with individual simulation-based learning. A simple computer-based randomization (1:1) was performed via a website [22] by an assistant not associated with the research. The evaluators and the statistical analyst were blinded to group assignment. Allocation concealment was achieved through opaque sealed envelopes. Participants were blinded to the research hypothesis. Due to the nature of this research, the instructor who conducted the simulation-based experience could not be blinded to group assignment.

The learning objective was to apply a physical therapy intervention based on exercise and education to a standardized patient with CLBP. Collaborative simulation-based learning (experimental group) was carried out according to the structure proposed by Guinez-Molinos et al [18]: (1) educational design: the creation of the evidence-based guide provided the theoretical framework for students to design, simulate, and evaluate collaborative scenarios; (2) collaborative design: 3 teams of 3 to 5 students were formed. Each team designed a clinical case based on the theme covered in class during the week. Students were also encouraged to write a brief clinical history. Additionally, they assigned specific roles within their simulation scenario. Through this method, weekly active participation was promoted among all students. Each week, the roles rotated among the students in each group; (3) simulation: the student guide explained the clinical simulation procedure to 2 students from another team and controlled the time. Observing students identified opportunities for improvement; (4) debriefing: unlike the protocol proposed by Guinez-Molinos et al [18], a plus-delta debriefing was used [23]. Conversely, the control group underwent individual simulation-based learning, following a traditional structure that used standardized patients. Details of each simulation-based learning protocol are presented in Multimedia Appendix 1.

The procedures for simulation-based learning were standardized into five stages to ensure validity and comparability between the groups:

Demographic data, such as age, sex assigned at birth (female or male), work, and number of courses, were collected in an online form. As potential confounder variables, the Academic Stressors Scale, the Academic Self-Efficacy Scale, and the R-14 Academic Resilience Scale were included in the online form.

All students were assessed individually before and after a 3-week simulation-based learning intervention (either collaborative or individual). The assessment consisted of a simulation-based scenario involving standardized patients. Clinical competency was defined as the primary outcome, while communication and therapeutic alliance served as secondary outcomes. Like other studies [24,25], students’ patient-centered communication skills were evaluated by standardized patients during the scenario; these patients had been previously trained to use the specific evaluation instrument. To minimize recall bias, the standardized patients completed their evaluations immediately following each encounter.

Consistent with previously published protocols, clinical competence and therapeutic alliance were evaluated via video recordings by independent physical therapists who were unfamiliar with the students and remained blinded to group assignments [26]. All students’ clinical performances were recorded using high-definition video and audio. To facilitate this, a camera equipped with an ambient microphone was positioned 3 m from the simulation scenario. This setup allowed for subsequent independent assessment by physical therapists with 5 years of experience in musculoskeletal rehabilitation. Prior to the assessment, these evaluators completed a 3-hour training session on the application of the instruments used to measure therapeutic alliance and clinical competence. This training involved evaluating 8 pilot video recordings, reviewing discrepancies, and receiving feedback from the principal investigator.

The pre- and post-intervention evaluations at the simulation-based scenario lasted 12 minutes, during which each student performed an intervention for a standardized patient experiencing a CLBP flare-up. Two clinical cases were developed to evaluate postintervention learning outcomes in both groups. A clinical case was used in the initial assessment, and another in the final assessment (Table S3 in Multimedia Appendix 2). The initial clinical characteristics were maintained to make the cases comparable, and only factors such as name, occupation, and family or occupational context were modified. To prepare the standardized patient and the physical space of the scenarios, the same training and review procedures were used as in the simulation-based learning group. In addition, in the training of the standardized patient, it was explained how to evaluate the communication of the students through a validated questionnaire.

This was evaluated with the Measurement Tool for Clinical Competencies in Physiotherapy (MTCCP), a valid and reliable (α=.98; intraclass correlation coefficient [ICC]=0.9) instrument to assess clinical competency in real and educational physiotherapy settings [27]. The MTCCP consists of 20 items, grouped into 2 subscales: professional behavior and clinical reasoning. The MTCCP uses a discrete measurement scale from 1 to 5. Those MTCCP items applicable to the simulated scenarios were selected for this research. Eight elements were considered to evaluate the video recordings (4 on the professional behavior subscale and 4 on the clinical reasoning subscale). Therefore, the maximum MTCCP score for this investigation was 40 points. A higher score indicates a better command of clinical competence, while lower scores indicate a worse command of clinical competence.

The sample size calculation was based on a power analysis “a priori compute required sample size” with an effect size (d=1.85) extracted from a previous and relevant study [30]. With an alpha of .05, power (1- β) of 95%, and Student t test, a minimum of 9 participants per group are needed. This sample estimate was increased by 20% after allowing for potential withdrawals, giving 11 participants in each group. The sample size calculation was performed using the G*Power 3.1.9.2 software (Department of Psychology at Heinrich Heine University Düsseldorf).

Descriptive statistics were used to describe the demographic and clinical characteristics of the participants. Continuous variables were the mean and SD, and categorical variables were the number and percentage.

Prior to inferential analyses, the tenability of the assumptions underlying the ANOVA model was systematically evaluated. The normality of each dependent variable was assessed within each group and time point using the Shapiro-Wilk test, supplemented by visual inspection of histograms and normal probability plots (Q-Q plots). The homogeneity of variances was assessed at each time point for all 3 outcomes using the Levene test. All contrasts were nonsignificant (P>.05). Given that all participants were assessed at both pre-intervention and post-intervention time points, data were analyzed using a 2 × 2 mixed-design ANOVA (also termed split-plot ANOVA), with group as the between-participants factor and time (pre vs post) as the within-participants (repeated measures) factor. Effect sizes were quantified using 2 complementary indices. Partial eta squared (η_p²) was computed for each source of variance in the mixed ANOVA model and interpreted according to the benchmarks: small (η_p²≈0.01), medium (η_p²≈0.06), and large (η_p² ≥0.14) [31]. Cohen d was additionally calculated for all planned post hoc comparisons as the standardized mean difference (ΔM/SDΔ for within-group paired comparisons; ΔM/SDpooled for between-group independent comparisons) and interpreted as small (d≈0.20), medium (d≈0.50), and large (d≥0.80) [32]. Given the exploratory nature of the pairwise comparisons and the priori enumerable set of 4 contrasts per outcome, post hoc analyses were conducted using paired-samples Student t tests for within-group (pre-to-post) comparisons and independent-samples Student t tests for between-group comparisons at each time point. To control the familywise type I error rate, all P values were adjusted using the Bonferroni correction (α’=.05/4=.0125 per comparison). The Bonferroni procedure was selected on the grounds of its conceptual transparency, conservative control of familywise error, and suitability for a small, prespecified set of contrasts [33]. Statistical analyses were performed using IBM SPSS Statistics for Windows (version 26; IBM Corp.).

This study was approved by the Research Ethics Committee of the Catholic University of Temuco (ID: 091302/22) and prospectively registered at OSF.io [34]. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki. All participants were informed of the study procedures, the voluntary nature of participation, and their right to withdraw at any time without penalty. Written informed consent was obtained from all participants prior to enrollment. Privacy and confidentiality were maintained throughout the study. The participants in this study did not receive any form of compensation for their participation.

The participants’ timeline can be seen in the flow diagram (Figure 2). A total of 31 students agreed to participate in this research. After randomization, 16 students were assigned to the individual group and 15 to the collaborative group. Eight students dropped out of the study: 1 due to vacation and 7 due to seasonal work. Therefore, 11 students in the individual group and 12 in the collaborative group completed the study. The characteristics of the participants are shown in Table 1. None of the students had previous experience in simulation-based learning with standardized patients. No differences were observed in physical therapy student characteristics at the beginning of the study. Furthermore, no differences were found between groups at baseline for clinical competence (P=.42), effective communication (P=.16), and therapeutic alliance (P=.44).

The inferential results from the mixed ANOVA, the detailed descriptive statistics for all outcomes, and their corresponding interaction trajectories are summarized in Tables 2 and 3 and illustrated in Figure 3. Additionally, the complete results for the Bonferroni-corrected post hoc pairwise comparisons are provided in Table S4 (Multimedia Appendix 2).

The mixed ANOVA revealed a large main effect of time on MTCCP scores (F_1,21=59.47, η_p²=0.739; P<.001), demonstrating substantial pre-to-post improvement in clinical competencies across both groups. The main effect of group was nonsignificant (F_1,21=0.595, η_p²=0.028; P=.45), and the group × time interaction did not reach statistical significance (F_1,21=0.264, η_p²=0.012; P=.61). However, the direction and magnitude of improvement were indicative of descriptive parity between conditions. Bonferroni-corrected post hoc analyses confirmed significant pre-to-post gains in both the collaborative group (t₁₁=−5.231, d=-1.510; P<.001 [corrected]) and the individual group (t₁₀=−5.979, d=1.803; P<.001 [corrected]). Between-group comparisons at preintervention (d=−0.346) and postintervention (d=−0.211) were small to moderate but nonsignificant after Bonferroni correction (both P_Bonf=1.00), reflecting a high degree of uncertainty due to the limited sample size.

For CAT scores, the mixed ANOVA yielded a negligible main effect of time (F_1,21=0.012, η_p²=0.001; P=.91), and the group × time interaction was nonsignificant (F_1,21=0.046, η_p²=0.002; P=.83). However, the main effect of group was characterized by a large effect size (F_1,21=3.906, η_p²=0.157; P=.06), despite missing the conventional threshold for statistical significance. Descriptively, the individual group scored marginally higher. Given that the large effect size (η_p²=0.157) alongside the nonsignificant P value, this trend strongly suggests a meaningful descriptive difference between the groups that cannot be conclusively confirmed or ruled out due to statistical underpowering.

The mixed ANOVA indicated a moderate-to-large main effect of time for WAI-O scores (F_1,21=8.744, η_p²=0.294; P=.008), reflecting a decrease in observer-rated working alliance from pre- to post-intervention across both conditions. The main effect of group (F_1,21=0.266, η_p²=0.013; P=.61) and the group × time interaction (F_1,21=0.546, η_p²=0.025; P=.47) were not statistically significant. Post hoc analyses revealed that the pre-to-post decline did not survive Bonferroni correction in either the collaborative group (t₁₁=1.637, d=−0.472; P_Bonf=.52) or the individual group (t₁₀=2.558, d=−0.771; P_Bonf=.11). Nevertheless, the moderate and large individual effect sizes reveal a descriptive reduction in alliance scores across both conditions, suggesting that true underlying changes may be obscured by statistical underpowering.

This study aimed to determine the effectiveness of collaborative versus individual simulation-based learning in developing clinical competence, effective communication, and therapeutic alliance among undergraduate physical therapy students during the care of standardized patients with CLBP flare-ups. Within the constraints of our sample size, our findings suggest that both simulation-based learning modalities led to within-group increases in clinical reasoning and professional behaviors, although no statistically significant differences were detected between the 2 modalities in clinical competence. However, observer-rated therapeutic alliance demonstrated a significant main effect of time, representing an overall decline across the intervention in both groups, while effective communication scores did not show statistical differences either between the groups or over time.

To the best of our knowledge, this study represents the first randomized experimental trial directly contrasting student-led collaborative scenario design against investigator-led individual simulation within physical therapy entry-to-practice curricula. The primary novelty of our work resides not merely in the logistical or financial advantages of collaborative simulation-based learning, but in providing preliminary empirical insight into a peer-assisted learning model for managing complex, psychosocially driven clinical conditions. Nevertheless, given the limited statistical power and the high uncertainty surrounding the resulting effect sizes, these findings should be considered exploratory rather than a definitive validation of the model’s magnitude of effect.

Several studies have corroborated the effectiveness of simulation-based learning for the development of physical therapy competencies involving practitioner skills, empathy, motivation, and attitudes [35-37]. While collaborative simulation-based learning has been applied in cardiology emergency scenarios [16], where the perceived usefulness among students and teachers was very high, its comparative effectiveness has not yet been established. Our results suggest that collaborative simulation-based learning may serve as an effective and comparable alternative to individual simulation-based learning with standardized patients. Collaborative simulation-based learning introduces a paradigm shift rooted in social constructivism and cognitive load theory [17,38,39]. By placing students at the center of the instructional design process, requiring them to collaboratively analyze the literature, cocreate clinical scripts, and enact the patient role, the cognitive load associated with complex clinical decision-making is functionally distributed [17,18]. To successfully construct an authentic CLBP flare-up scenario, student dyads must actively negotiate and internalize what a patient experiencing centralized pain, catastrophizing, and fear-avoidance actually undergoes [5,6]. This metacognitive phase prior to execution could explain why the collaborative group achieved clinical competence scores similar to those of the individual group.

Regarding effective communication and the therapeutic alliance, both variables showed no significant differences between groups at the end of the study. While communication remained relatively stable over time, surprisingly, the therapeutic alliance decreased significantly between the pre- and post-intervention periods. Rather than reflecting a failure in teaching, this systematic reduction in the quality of the therapist-patient bond reflects a well-documented phenomenon in medical education, the cognitive rigidity of the novice physical therapist when faced with clinical complexity [40-42]. A flare-up of CLBP generates intense emotional distress and frustration in patients, which in turn complicates the relational context during the clinical encounter [5,6]. High-fidelity and complex standardized patient simulations can potentially induce significant anxiety and negative emotions in undergraduate students. These emotional responses may consume working-memory resources by increasing extraneous cognitive load, potentially hindering learning if total cognitive load exceeds working memory capacity [43,44]. Consequently, to preserve performance, students frequently retreat into a defensive, more technical, and highly prescriptive biomedical mode, focusing strictly on the physical execution of manual techniques or exercise parameters while neglecting affective validation, active listening, and the encouragement of positive expectations [45,46]. Given that an enhanced therapeutic alliance and effective communication are contextual factors capable of significantly reducing pain intensity [47], this procedural limitation represents a critical clinical error.

As not all students can participate in the clinical experience of meeting and interacting with this spectrum of patients, simulating a high-fidelity scenario provides an opportunity to practice and achieve clinical competencies in a safe learning environment. An example of a complex scenario is the management of a flare-up in patients with CLBP, as it impacts major elements of quality of life and involves multidimensional factors, such as a sense of disablement and shifts in mood or emotions, that the student must manage [5,6,48]. Active therapy with a patient-centered approach can modulate the patient’s outcome through communication strategies such as active listening, empathy, building positive expectations, and providing reassuring information [45]. Therefore, promoting a therapeutic alliance and effective communication, while taking into account contextual factors, can be implemented as a key clinical education strategy within the university curriculum of health care professionals [45]. Through either collaborative or individual simulation-based learning, real-life experiences are represented in a guided and interactive manner; however, a collaborative approach may reduce overall costs of simulation, specifically with script development and the cost of actors [49].

To integrate collaborative simulation-based learning into physical therapy or other health sciences curricula, several guidelines are recommended. First, it is necessary to identify the course, learning outcomes, and competencies to be developed to determine how many collaborative SBEs will be integrated. Second, the instructor must provide reading material that serves as the baseline input for students to create their clinical cases and scenarios. Third, students should be informed in advance of the simulation modality and their specific responsibilities. Fourth, during the collaborative simulation-based learning design phase, the instructor must facilitate the creative process by resolving doubts, defining roles, and reviewing the clinical coherence of the scenario and case against the intended learning outcome (eg, the general characteristics of the patient, presence of a companion, signs, symptoms, psychosocial factors, equipment, and spatial layout). Fifth, the execution of the scenarios is the sole responsibility of the designated student team, requiring members to fulfill the roles of patient, companion, timekeeper, and performance evaluator. Sixth, the debriefing phase can be adapted according to the methodology that best suits the instructor guiding the process.

From an educational curriculum standpoint, these preliminary findings suggest that collaborative simulation-based learning should not be restricted to an introductory method or a contingency plan for low-resource settings. Instead, it should be strategically integrated as a scaffolded instructional methodology. We recommend implementing collaborative peer-designed simulation during early entry-to-practice phases to build solid clinical reasoning and communicative baselines in psychologically safe, peer-supported environments. Once these distributed cognitive frameworks are consolidated, students can progressively transition toward individual encounters with professional standardized patients, culminating in immersion with real, complex clinical populations.

To the best of our knowledge, this is the first comparative experimental study directly evaluating the effectiveness of collaborative versus individual simulation-based learning. Following a previously defined and easily reproducible protocol, we trained standardized patients for both the evaluation phases and the simulation-based learning interventions. The assessment sessions were video recorded to prevent students from experiencing heightened stress due to the physical presence of an in-room evaluator. Finally, the evaluation incorporated the dual perspectives of the standardized patient and a blinded external evaluator. However, our findings should be interpreted with caution due to several limitations. Although effective communication and the therapeutic alliance were evaluated, these relational skills were not explicitly included as formal content or core competencies to be developed during SBEs. Furthermore, while we performed a priori sample size calculation based on the existing literature, our final sample size limited the statistical power of the study. In underpowered educational research, these constraints can lead to overinflated effect sizes, a phenomenon that introduces a large amount of uncertainty regarding the true magnitude of the intervention’s impact. Consequently, the unusually high effect sizes observed in our results must be interpreted with caution, as they may restrict the generalizability of the findings to broader physical therapy curricula. Future studies should incorporate larger and more diverse sample sizes, incorporate explicit post hoc power analyses, assessments of student stress and anxiety, and cost-effectiveness analyses, alongside qualitative evaluations of the collaborative simulation experience. Because this study included only physical therapy students, further research is warranted across other health professions, including interprofessional approaches.

In a complex clinical management context, both collaborative and individual simulation-based learning led to increases in physical therapy clinical competencies, with no statistically significant differences detected between the 2 instructional modalities. However, there were no differences between the groups for effective communication or therapeutic alliance. Given the high magnitude of improvement observed across both modalities, detecting a statistically significant differential effect between them would require larger sample sizes. Further research is warranted to assess the efficacy of these modalities in diverse clinical settings.