A virtual patient-based training program to enhance clinical decision-making competencies in general practice residents: mixed-methods study

Abstract

Background

Clinical decision-making (CDM) represents a fundamental competency for general practice (GP) residents, yet traditional training approaches often provide limited opportunities for deliberate practice. Virtual patient (VP) platforms have emerged as promising educational tools that offer standardized, repeatable, and risk-free environments for developing clinical reasoning competency. However, empirical evidence examining the effectiveness of structured VP training programs in enhancing CDM competencies among GP residents remains limited, particularly regarding the mechanisms through which such training facilitates competency improvement. This study aimed to evaluate the effectiveness of a VP training program in enhancing CDM competencies among GP residents, and to explore their perceptions of the program’s value and barriers in developing these competencies.

Methods

An explanatory sequential mixed-methods design was employed, comprising a quantitative pre-post assessment followed by qualitative exploration. Over 12 weeks, 50 GP residents participated in a curriculum involving weekly exposure to two VP cases for self-paced practice, supplemented by weekly facilitated workshops to discuss these cases. Quantitative pre-post assessments included OSCEs (history-taking, physical examination, clinical reasoning, medical record writing) and a theoretical test. Subsequently, eight focus group interviews were conducted to explore their learning experiences and perceived barriers to improvement of CDM competencies. Data were analyzed using paired t-tests for quantitative data and thematic analysis for qualitative data.

Results

Quantitative assessment of the 50 participants (mean age 25.5 years; 70% female) demonstrated statistically significant but differential improvements in clinical competencies. Following the VP training program, GP residents demonstrated significant improvements in clinical knowledge (mean increase 6.7 points, 95% CI [4.2, 9.1]), history-taking (mean diff.=10.2, 95% CI [7.6, 12.8]), and clinical reasoning (mean diff.=11.2, 95% CI [8.8, 13.6]). Conversely, observed improvements in physical examination (mean diff.=2.4, 95% CI [-0.1, 4.8]) and medical record writing (mean diff.=1.2, 95% CI [-0.1, 2.5]) were not statistically significant. Qualitative findings suggested that residents attributed their CDM competency development to the platform’s structured design, which cultivated systematic diagnostic reasoning, adaptive competence in uncertainty, and promoted reflective learning. Technological fidelity limitations in physical exam simulation, restrictive documentation interfaces, challenges in skill transfer to real practice, and variable workshop effectiveness were identified as core barriers to the implementation of VP training program.

Conclusion

A structured VP curriculum embedded in GP residency was associated with improved clinical knowledge, history-taking, and clinical reasoning, while showing limited effects on physical examination and medical record writing. Future development should prioritize advancing technological realism in these areas, optimizing curricular integration through facilitated debriefing, and investigating the long-term transfer and retention of acquired competencies into clinical practice.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12909-026-09607-8.

Introduction

Clinical decision-making forms a cornerstone of general practice. Physicians in primary care routinely encounter undifferentiated symptoms and a wide range of conditions [1]. Accurate diagnosis and treatment depend on the ability to synthesise patient history, physical signs, and test results [2]. Teaching this complex competency at the postgraduate level remains difficult. Traditional bedside instruction relies on unpredictable clinical cases. Practising on real patients carries inherent risks. Feedback from supervisors often arrives late or lacks consistency [3, 4]. A shortage of clinical faculty further limits individualised training [5]. Digital teaching methods provide a concrete solution. They offer standardised, repeatable and safe learning environments.

As promising digital tools [6], virtual patients (VPs) are interactive clinical simulations that actively engage learners in taking histories, ordering investigations, and formulating management plans, all guided by immediate automated feedback [7–9]. Their primary advantage lies in facilitating deliberate practice within a safe, risk-free environment [8]. Substantial evidence shows VPs effectively complement traditional education through active learning [10]. Compared to conventional methods, they demonstrate equal or superior efficacy in enhancing overall clinical knowledge and are particularly valuable for cultivating context-dependent reasoning skills, such as data gathering and diagnostic hypothesis generation [11].

Within the domain of general practice training, the application of VPs is gaining recognition for its potential to simulate the breadth and uncertainty characteristic of primary care. A few studies have begun to explore the use of VPs for teaching GP residents, often focusing on chronic disease management or clinical reasoning [12, 13]. These investigations generally affirm the acceptability and perceived utility of VPs among GP residents. However, existing studies primarily focus on short-term interventions or discrete skills, lacking evidence on the longitudinal integration of VPs within a complete residency curriculum [14]. Crucially, the mechanisms through which sustained VP training cultivates complex CDM competencies remain poorly understood. Furthermore, barriers that determine the successful implementation of such programs in real-world training settings remain underexplored [15].

To address this gap, the curriculum integrates repeated VP practice with structured expert debriefing. Guided by constructive alignment [16], it directly aligns targeted competencies, including clinical reasoning and management planning, with active case engagement and tailored assessments. Crucially, the post-VP debriefing allows residents to compare reasoning pathways, correct misconceptions, and translate case-specific feedback into generalizable decision-making strategies.

This study evaluated a 12-week VP-based curriculum for GP residents using a mixed-methods curriculum evaluation approach. It sought to determine the program’s effectiveness in enhancing CDM competencies and to elucidate the barriers to its implementation. The investigation provides integrated evidence on both the quantitative impact and qualitative realities of sustained VP integration in GP residency training.

Methods

Study design

This investigation employed an explanatory sequential mixed-methods study design, as shown in Fig. 1 [17]. The quantitative phase consisted of a pre-post intervention study, evaluating the impact of a structured VP training program on the CDM competencies of GP residents. Objective Structured Clinical Examination (OSCE) scores and theoretical test results were collected before and after the program. Subsequently, the qualitative phase was conducted, comprising a series of focus group interviews following the post-test assessments. The integration of quantitative and qualitative data occurred during the interpretation phase to provide a comprehensive understanding of the program’s effectiveness and implementation.

Fig. 1.

Overview of the explanatory sequential mixed-methods study procedure

Ethical consideration

This study was approved by the Ethics Committee of the Second Affiliated Hospital of Zhejiang University School of Medicine [Approval No: (2024) Ethics Review for Research No. (1048)], in accordance with the Declaration of Helsinki [18]. All participants provided written informed consent prior to their involvement in the study. The consent process explicitly detailed the study’s purpose, procedures, potential risks and benefits, and the assurance of confidentiality. Participants were informed of their right to withdraw from the study at any time without penalty to their training or professional standing. For participants who chose to withdraw during the training program, data collected up to the point of withdrawal were retained for analysis unless they specifically requested its deletion.

Setting and participants

This study was conducted within the GP Residency Training Base at the Second Affiliated Hospital of Zhejiang University School of Medicine. The base operates a standardized three-year training program for resident physicians, structured in accordance with the national requirements set by the Chinese Medical Doctor Association. The curriculum mandates rotational postings across 21 clinical departments, including General Practice, Internal Medicine, surgery, pediatrics, and gynecology, etc., supplemented by community healthcare training. The duration of these rotations varies from two weeks to six months, ensuring comprehensive exposure to a wide spectrum of medical disciplines and patient care settings.

The study participants comprised the entire cohort of 50 general practice residents enrolled in the training program during the 2024–2025 academic year. Participation in the VP training program was integrated into the standard curriculum. All 50 residents consented to participate in the study component. No financial compensation was provided for participation in the research.

Virtual patient platform

The training intervention was delivered using the “CureFun” VP platform, an interactive web-based system accessible via both desktop and mobile devices [19]. This platform hosts a library of over 1,500 clinical cases covering a wide range of medical conditions. For this training program, a panel of clinical experts developed 24 customized VP cases representing common scenarios in GP’s ambulatory care. The platform’s editing functionality was then utilized to build these scenarios, ensuring they were tailored to align with the learning objectives for developing GP residents’ CDM competencies.

The platform simulates authentic clinical encounters. Learners interact with the virtual patient by clicking pre-formulated questions from a menu to obtain the patient’s history, and conduct a physical examination by selecting appropriate maneuvers from a list (Figure S1 in Multimedia Appendix 1). They then proceed to order and interpret auxiliary examination, and ultimately formulate diagnoses and treatment plans (Figure S2 in Multimedia Appendix 1). A key feature of the platform is its integrated assessment system, which provides multi-dimensional scoring and immediate, automated feedback. This feedback provides multi-dimensional scoring of learners’ clinical actions across six key domains: (1) appropriateness of history-taking, (2) appropriateness of physical examination, (3) appropriateness of auxiliary examination selection, (4) appropriateness of treatment planning, (5) diagnostic accuracy, and (6) cost-effectiveness awareness (Figure S3 in Multimedia Appendix 1). Additionally, the platform provides educators with a comprehensive dashboard interface, allowing them to review and compare all learners’ clinical actions and assessment for each virtual patient case (Figure S4 in Supplementary file 1).

VP training program

A structured VP training program was integrated into the standard residency curriculum. The program followed a consistent weekly cycle: on the Tuesday of each week, two new VP cases from the were released simultaneously through the platform. These cases were carefully selected to represent either complementary clinical presentations (e.g., chest pain and dyspnea) or contrasting diagnostic challenges (e.g., acute vs. chronic headache) to facilitate comparative CDM. Residents were required to complete both cases independently. The platform allowed unlimited attempts during this period, enabling learners to refine their diagnostic approach through repeated practice. Each case required up to 30 min to complete, simulating a focused clinical encounter. The system tracked all learner interactions, including history questions asked, physical examination maneuvers selected, additional investigations ordered, diagnosis made, and treatment plans formulated. Upon case completion, residents received immediate automated feedback evaluating their clinical performance.

The curriculum design deliberately integrated a post-VP in-person expert debrief. While Virtual Patients are highly effective in developing core clinical competencies, prior literature indicates that standalone screen-based simulations often lack the capacity for nuanced feedback and deep cognitive reflection [20, 21]. Furthermore, students engaging in virtual simulations can experience cognitive overload or harbor uncorrected misconceptions without guided facilitation [22]. To address this gap, an in-person expert debrief was incorporated.

Upon completion of each pair of cases, a facilitated 2-hour workshop convened weekly on Mondays. During these sessions, two residents were selected to present their CDM process for the cases with prepared PowerPoint slides, including their case summary, differential diagnoses, key findings, and management plans. Following each presentation, two attending physicians specializing in general practice provided structured feedback and expert commentary on the clinical approach. A senior GP faculty member facilitated the workshops, guiding all residents through a reflective discussion that compared different diagnostic strategies, explored alternative management options, and highlighted key learning points from the cases. This design created a continuous learning cycle of individual practice, peer learning, and expert feedback, reinforcing the development of CDM competencies throughout the training period.

CDM competencies evaluation

GP residents’ CDM competencies was evaluated using an assessment framework administered at baseline (September 2024) and upon program completion (December 2024). The evaluation consisted of an Objective Structured Clinical Examination (OSCE) and a theoretical examination. Two parallel but non-identical versions of both the OSCE stations and the theoretical examination were developed to minimize the potential for recall bias between the pre- and post-intervention assessments. The content, complexity, and clinical domains covered in these parallel versions were rigorously matched by a panel of clinical and assessment experts to ensure equivalent difficulty and comparable measurement of the targeted competencies.

Administered as a 100-point computer-based test, the theoretical examination comprised five formats of single-best-answer multiple-choice questions (A1–A4 and B1) designed to assess core medical knowledge and clinical application. These formats systematically ranged from basic factual recall and standard matching to the progressive analysis of both static and dynamically evolving clinical case scenarios. The OSCE component consisted of four stations aligned with the standardized competency assessment framework for residency completion in our jurisdiction: history-taking, physical examination, clinical reasoning, and medical record writing. Each station was scored on a 100-point scale using the official, validated assessment checklists issued by the Zhejiang Provincial Health Commission for residency clinical competency evaluation, ensuring that the measurement tools and processes directly reflected training objectives and evaluation standards. Two independent GP experts, both formally certified as residency clinical competency examination assessors by the Zhejiang Provincial Health Commission, evaluated each resident’s performance at every station. The final score for each station was calculated as the average of the two assessors’ ratings, following the standard protocol for OSCE administration.

Focus group interviews

We conducted eight semi-structured focus group interviews in January 2025 to explore GP residents’ learning experiences and perceived barriers to CDM development. A total of 46 residents participated in these sessions, with 4–7 participants per group, while four residents were unable to attend due to scheduling conflicts. Informed by the initial quantitative results and a literature review, the interview guide was designed with broad, open-ended questions and refined through expert feedback and team discussions. During the actual sessions, however, facilitators dynamically referenced the quantitative data to probe deeper into participants’ underlying reasoning (Table S1 in Supplementary file 2).

Each 90–120 min session was co-facilitated by two individuals: the primary facilitator (YC), an experienced qualitative researcher who had no prior professional or supervisory relationship with any of the residents, thereby ensuring methodological neutrality and minimizing social desirability bias. The second facilitator (LW) was a general practice trainer from the residency program, who assisted with participant coordination, managed logistical arrangements, and documented non-verbal cues and contextual observations during the sessions. The primary facilitator employed open-ended questioning techniques to encourage participants to freely share their perspectives while ensuring comprehensive coverage of key topics. All interviews were conducted in a private conference room, audio-recorded, and subsequently transcribed verbatim for analysis, with any identifying information removed during transcription to ensure participant confidentiality.

Data analysis and integration

Descriptive statistics were presented as mean ± standard deviation for normally distributed continuous variables, and as counts and percentages for categorical variables. The normality of score distributions was assessed using the Shapiro-Wilk test. To evaluate the impact of the training program, we compared pre- and post-intervention scores using paired t-tests for normally distributed data, with results reported as mean differences, 95% confidence intervals, and Cohen’s d effect sizes. Quantitative data were analyzed using SPSS Statistics (Version 22.0).

Qualitative data underwent thematic analysis following Braun and Clarke’s six-phase framework [23]. Two researchers familiarized themselves with the transcripts and generated initial codes. Subsequently, a codebook was developed, containing explicit code definitions and representative exemplars. The two researchers then applied this codebook to the complete dataset using MAXQDA V.20.4.2 (VERBI Software, Berlin). Through an iterative process and the constant comparative approach, these codes were collaboratively organized into potential themes and subthemes, which were continually refined through research team discussions. Emerging themes were checked against raw data to ensure they accurately represented participants’ perspectives.

The integration of data occurred during the interpretation phase of this explanatory sequential mixed-methods study. We employed a joint display technique to facilitate explicit comparison and connection between the two datasets. The primary quantitative results were tabulated. Alongside, the corresponding qualitative themes and illustrative quotations that either explained, expanded, or contradicted those results were presented in a parallel column. The qualitative findings were used to provide an explanation for the quantitative outcomes, particularly to interpret the mechanisms behind improvements and to elucidate the reasons for non-significant changes in specific competency domains. This integrative analysis aimed to generate a more comprehensive understanding of the program’s effectiveness and implementation barriers.

Results

Participant characteristics

All fifty enrolled general practice residents completed the 12-week training program and both assessments. The group consisted of 35 females (70%) with a mean age of 25.5 ± 2.5 years (range: 23–34 years). The majority of participants (88%) held bachelor’s degrees as their highest qualification. The majority of participants (80%) were employed by community health centers, while 4% were not currently affiliated with any medical institution (Table 1).

Table 1.

Characteristics of study participants (N = 50)

Characteristic	Value
Age (years)
Mean ± SD	25.5 ± 2.5
Range	23–34
Gender, n (%)
Female	35 (70.0)
Male	15 (30.0)
Grade level
First-year	16 (32.0)
Second-year	17 (34.0)
Third-year	17 (34.0)
Highest Education, n (%)
Bachelor’s degree	44 (88.0)
Master’s degree	2 (4.0)
Doctoral degree	4 (8.0)
Work Institution, n (%)
Community health center	40 (80.0)
Comprehensive hospital	8 (16.0)
Unaffiliated	2 (4.0)

Clinical knowledge assessment

Clinical knowledge scores demonstrated improvement following the VP training program. The mean score increased 6.7 points from baseline (68.1 ± 6.4) to post-intervention (74.8 ± 8.0). This improvement reached statistical significance [95%CI (4.2, 9.1), p< .001], suggesting a large magnitude of change in theoretical knowledge as measured by the standardized examination (Cohen’s d = 0.76).

OSCE performance outcomes

Analysis of OSCE station scores revealed differential improvements in clinical competencies following the VP training intervention. As presented in Fig. 2, residents demonstrated statistically significant improvements in history-taking (mean difference = 10.2, 95% CI [7.6, 12.8], Cohen’s d = 1.11) and clinical reasoning (mean difference = 11.2, 95% CI [8.8, 13.6], Cohen’s d = 1.32). Improvements in physical examination (mean difference = 2.4, 95% CI [-0.1, 4.8], Cohen’s d = 0.28) and medical record writing (mean difference = 1.2, 95% CI [-0.1, 2.5], Cohen’s d = 0.26) were not statistically significant, with small effect sizes observed. The detailed scores stratified by training year and for the overall cohort are provided in Table S2 (see Supplementary file 3).

Fig. 2.

Comparison of CDM competency scores before and after the program. Footnote: ***, P < 0.001. Error bars represent 95% confidence intervals

Perceived impact of VP training program on CDM competency development

Enforcing a structured decision-making approach

The focus groups revealed that the VP training was instrumental in fostering a more structured clinical reasoning process among GP residents. The virtual platform guided residents through a sequential pathway. Residents formulated active hypotheses prior to treatment selection. This sequential progression functioned as a cognitive framework which cultivated metacognition.

Participants perceived the immediate, criterion‑referenced feedback as a key mechanism for consolidating this cognitive change. Residents refined their clinical judgment through this iterative practice, eventually transforming these cognitive steps into durable habits.

Cultivating adaptive competence for clinical uncertainty

Residents perceived the VP training as enhancing their cognitive flexibility and ability to manage clinical uncertainty through continuous exposure to diverse medical scenarios. This experiential learning helped them refine internal mental models, which are cognitive scripts or heuristics developed through training to enable efficient clinical reasoning [24]. Developing robust mental models through repeated simulation can enhance clinical efficiency by enabling more rapid diagnostic reasoning. The platform’s branching narratives demonstrated the consequences of diagnostic and therapeutic choices in a safe environment, cultivating an adaptive mindset.

Moreover, residents were required to justify management decisions for complex, multimorbid cases, moving beyond single-disease frameworks. This process fostered integrated care plans that account for competing priorities and polypharmacy risks, strengthening adaptive competence.

Promoting reflective learning

VP training shifted residents from automatic responses to conscious evaluation of their clinical reasoning through multi-dimensional feedback on effectiveness, safety, and efficiency. This feedback provided an objective benchmark for calibration, enabling learners to continuously reflect on and adjust their decisions.

Weekly facilitated debriefing sessions transformed isolated VP exercises into collaborative learning experiences. By verbalizing their clinical reasoning to peers and experts and hearing alternative diagnostic pathways, residents were exposed to a broader spectrum of cognitive approaches. This process prevented the entrenchment of errors, helped solidify correct mental models, and ultimately deepened their clinical decision-making competency.

Perceived barriers to implementing VP training program

Limited physical examination fidelity

A common barrier to the development of CDM competencies, as identified by residents, pertained to the technological limitations and insufficient fidelity of the VP platform in simulating key aspects of the physical examination. While the system was effective for training diagnostic reasoning, its inability to replicate nuanced physical signs was consistently reported as a critical shortcoming.

Restrictive clinical documentation interface

A notable unintended consequence of the VP training was its negative impact on the development of proficient medical record writing skills. The virtual platform shaped documentation habits through its menu-driven interface. Residents selected specific findings from digital lists. This structured format fostered a fragmented approach to documentation. The system highlighted mechanical data entry. Residents characterized the resulting records as a sequence of procedural clicks which separated the writing task from the articulation of diagnostic thought.

Difficulties in transferring competencies to real practice

A critical barrier identified by focus groups was the challenge in seamlessly transferring CDM competencies honed in the virtual environment to actual clinical practice. Residents reported difficulties in applying their structured VP-trained approach amidst the time pressures, interruptions of live patient care and interpersonal nuances.

Inadequate peer learning in workshops

Several participants indicated that the quality of workshops was highly dependent on facilitators’ teaching skill, and their inadequate preparation for this new pedagogical role may result in variable session quality. Some sessions became dominated by a few vocal participants or devolved into simple case review rather than deep reflective discussion, limiting their impact on collective CDM development (Table 2).

Table 2.

Qualitative themes and representative quotes

Theme	Sub-theme		Representative quotes
Perceived impact	Enforcing a structured decision-making approach	Systematic diagnostic reasoning	Before, I would often jump to the most common diagnosis. Now, with the VPs, I’ve developed a mental checklist. I actively think, ‘What are the top three possibilities for this chest pain, and what specific history or exam findings would rule each in or out?’
		Immediate feedback consolidation	The system forces you to follow a clinical pathway, which has rewired my approach to be more efficient and grounded. Getting immediate feedback on why a certain path was inefficient was sometimes frustrating but ultimately very effective for learning.
	Cultivating adaptive competence for clinical uncertainty	Cognitive flexibility	During our clinical rotations, sometimes we might go for months without seeing a single classic case worth revisiting for deeper learning, but with the VPs, I was navigating a challenging scenario every week. It trained me to be comfortable with not having all the answers immediately.
		Complex, multimorbid case management	The case with the older diabetic patient with heart failure was a revelation. It wasn’t just about lowering blood sugar, so I had to think about which medication would be safest for her kidneys and heart. It pushed my management planning to a whole new level of complexity.
	Promoting reflective learning	Multi-dimensional feedback and calibration	The feedback was important, I think. I once ordered what I thought was the perfect evidence-based treatment, but the system flagged that I hadn’t considered the patient’s financial situation. It taught me that the medically correct decision isn’t always the right decision for that person.
		Facilitated debriefing as critical	Completing the VP case was just the first step. The real learning happened in the workshop when I heard how another resident had approached this case completely differently, and the expert explained why both paths could be valid depending on your initial hypothesis. It opened my mind to flexibility in thinking.
Perceived barriers	Limited physical examination fidelity		I could list ‘auscultate the heart’ in the system, but the feedback was just a binary correct/incorrect based on if I chose the action. It didn’t simulate the experience of actually hearing a murmur and having to distinguish its timing and location.
	Restrictive clinical documentation interface		When I returned to writing a real clinic note, I found myself struggling to string the individual findings into a flowing history of present illness. I had the data, but sometimes I’d lost the practice of telling the patient’s story.
			In a real chart, my assessment and plan explain my reasoning. But in the VP system, the reasoning was just the sequence of clicks I made. There was no space to articulate why I was ruling out a particular diagnosis.
	Difficulties in transferring competencies to real practice		In the simulation, I could take 20 to 30 min to methodically work through a case. But in the clinic, I have 10–8 min per patient and a waiting room full of people. The pressure makes you fall back on old habits.
	Inadequate peer learning in workshops		In a few sessions, the discussion was hijacked by two or three very confident residents. The rest of us just listened. We missed the chance to hear from the quieter ones who might have had a different, and perhaps more insightful view.

Integration of quantitative and qualitative findings

The explanatory sequential mixed-methods design provided comprehensive insights into the mechanisms underlying the differential effectiveness of the VP training program (Table 3). Quantitatively, participation in the training program was associated with statistically significant improvements in history-taking, clinical reasoning, and clinical knowledge scores, whereas changes in physical examination and medical record writing scores were marginal and non-significant. Qualitatively, participants attributed the statistically significant skill enhancements to the platform’s structured design, which cultivated systematic diagnostic reasoning, adaptive competence in uncertainty, and reflective learning. Conversely, participants identified the platform’s inability to simulate nuanced physical examination findings, its restrictive, non-narrative documentation interface and inadequate peer learning in workshops as core barriers, which elucidate the limited quantitative progress observed in these competencies. These integrated findings confirm where and how the VP training is effective, and clarify why its impact remains constrained in specific competency areas.

Table 3.

Joint display of quantitative and qualitative results

Quantitative results	Qualitative findings	Relationship	Integrated interpretation
Significant improvement in clinical reasoning and history-taking OSCE scores.	Residents identified the structured design of the VP platform and the immediate, criterion-referenced feedback as key mechanisms that fostered systematic thinking and refined their diagnostic/treatment approach.	Confirmation & Explanation	The qualitative data reveals the specific cognitive mechanisms (e.g., structured reasoning, metacognition) that were cultivated by the VP training, explaining how the substantial gains in clinical reasoning and history-taking scores were observed after the program.
Significant improvement in clinical knowledge scores.	Residents described how the repeated, diverse case exposure and the need to justify clinical decisions in complex scenarios deepened their understanding and application of medical knowledge.	Confirmation & Expansion	The qualitative data suggests that the knowledge gain was not merely factual recall but involved applied and integrated learning, complementing the quantitative result by indicating that the VP training may enhance the contextualized use of knowledge.
Non-significant, small improvements in physical examination OSCE scores.	Residents reported the technological limitation and low fidelity of the VP platform in simulating nuanced physical signs.	Explanation	The VP platform and workshop did not adequately train or assess the sensory and procedural clinical skills central to physical examination, which may explain the non-significant improvement in this competency.
Non-significant, small improvement in medical record writing OSCE scores.	Participants criticized the restrictive, menu-driven documentation interface, stating it fostered fragmented data entry and actively hindered the development of narrative, synthesis-based clinical note-writing skills.	Explanation	The qualitative reports of a limited function of VP platform in medical recording writing align with the stagnant quantitative scores. The platform’s interface was mismatched with the competency being measured, which may explain the lack of progress in medical record writing.

Discussion

This explanatory sequential mixed-methods study examined the association between a structured VP training program and the development of CDM competencies among general practice residents. While integrating quantitative and qualitative data revealed improvements in specific CDM domains and elucidated potential educational mechanisms, the single-group, pre-post design precludes definitive causal attribution. Observed gains may reflect concurrent clinical maturation during residency. Thus, our findings should be interpreted as demonstrating an association between VP training program and enhanced CDM competency within the existing residency training context.

Our finding that VP training improved clinical knowledge and history‑taking aligns with a broad body of evidence showing that VP‑based interventions enhance CDM competencies across health professions [25–31]. Residents perceived the platform as enforcing structured decision‑making, which matches qualitative reports that VPs promote logical reasoning, safe trial‑and‑error practice, and reflection [32]. Criterion‑referenced feedback consolidated metacognitive skills by clarifying the links between clinical actions and their consequences [33, 34]. Repeated exposure to diverse cases cultivated adaptive competence under uncertainty, helping learners adjust mental models and move beyond single‑disease algorithms [35]. However, these benefits did not distribute evenly across training levels. The curriculum delivered the greatest benefit for first-year residents, while third-year residents showed minimal changes. This pattern suggests that VP training works best when introduced early in residency, before clinical experience solidifies automatic reasoning. Senior residents may require more advanced simulation formats, such as time-pressured scenarios, to further challenge their established thinking [36]. Even with these positive outcomes, the VP platform alone is a necessary but insufficient component. Its educational power depends on the surrounding pedagogical wrap‑around, including feedback, reflection, and peer dialogue [37, 38].

Rather than merely replicating prior findings, this study makes a distinct contribution through its intentional pairing of repeated VP case practice with weekly facilitated workshops within a longitudinal 12-week curriculum for GP residents. This design was guided by constructive alignment principles, ensuring that targeted competencies were directly matched with learning activities and assessments. The curriculum demonstrated selective competency enhancement, with the largest gains observed in history-taking, clinical reasoning, and clinical knowledge, while physical examination and medical record writing showed limited improvement. This differential pattern is noteworthy because it highlights which cognitive competencies are most amenable to VP-based training and which domains require complementary pedagogical approaches. The significant improvements in clinical reasoning and history-taking align with the CanMEDS-Family Medicine 2017 framework, which identifies structured clinical assessment and diagnostic reasoning as core competencies for family physicians [39], and with the ACGME Family Medicine Milestones that emphasize progressive development of prioritized differential diagnosis and clinical reasoning skills [40]. Our results suggest that the curriculum effectively addresses these core cognitive competencies expected of GP trainees in North American training systems.

The facilitated workshops represented a critical pedagogical component, transforming individual VP exercises into collaborative learning opportunities through structured debriefing. However, the variable quality of peer learning reported by residents highlights the need for evidence-based workshop facilitation. Research on simulation debriefing has established that structured approaches such as the PEARLS framework, which integrates learner self-assessment, focused discussion, and directive feedback, can significantly improve debriefing quality and educational outcomes [41]. Facilitator training in these structured debriefing methods, combined with strategies to ensure equitable participation across all learners rather than domination by vocal individuals, may address the workshop barriers identified in this study [22, 42]. These approaches could be integrated into faculty development programs to enhance the effectiveness of VP-based workshop sessions. Integrating large language models into VP platforms may address current limitations by enabling natural language interaction for history-taking, dynamic scenario adaptation, and more nuanced feedback generation, potentially expanding the range of competencies that can be effectively trained through virtual simulation [26].

Notebly, the CureFun platform is widely used in Chinese medical education, offering desktop and mobile access with automated feedback. Its comprehensive feature set exceeds earlier platforms with limited case libraries [43]. However, residents select physical examination items from a menu rather than performing actual maneuvers, which likely contributed to weak score gains. The menu‑driven documentation interface similarly explains poor medical record writing. Other VP systems, such as DxR Clinician or Body Interact, provide audio clips for heart and lung sounds, natural language processing for open‑ended history taking, and large language models for conversational dynamics [44, 45]. CureFun holds clear advantages in local adaptation and system completeness. Future upgrades could prioritise the inclusion of multimedia physical signs and open‑text documentation fields.

Several limitations of this study warrant consideration. First, although the single-center design and the relatively small absolute sample size may limit generalizability, it is worth emphasizing that our sample included all GP residents from the training base (N = 50, 100% retention), and the single-group pre-post design allowed us to capture the entire target population without sampling bias. With a median sample size of 25–30 in medical education simulation research [46], our sample size is well within the acceptable range, supporting its methodological. Second, the absence of a parallel control group prevents isolating the VP intervention’s effect from concurrent clinical maturation. Observed improvements should be interpreted as associations rather than definitive causal outcomes. Third, OSCE performance may not fully capture real‑world CDM complexity. Fourth, the sequential design introduced a time delay between quantitative post‑test and qualitative collection, and the interview guide focused on pre‑identified outcomes, potentially missing unexpected themes. Fifth, social desirability bias may have affected focus group responses, though facilitators emphasized anonymity. Finally, the follow‑up period does not allow assessment of long‑term competency retention, so longitudinal studies are needed.

Conclusion

This study demonstrates that integrating a structured VP training program into GP residency is associated with enhanced CDM competencies, particularly in clinical reasoning, history-taking, and clinical knowledge. The mixed-methods evidence confirms that these improvements are linked to the program’s capacity to foster systematic diagnostic reasoning and reflective practice through structured simulation and facilitated debriefing. Simultaneously, the research identifies specific constraints: current technological limitations in simulating physical examinations and supporting clinical documentation, along with variable implementation quality in workshop facilitation and peer learning dynamics. Future efforts should therefore prioritize advancing technological realism in these constrained domains, optimizing curricular integration through evidence-based facilitation strategies, and systematically investigating the long-term retention and clinical transfer of acquired competencies.

Abbreviations

VP

Virtual patient

GP

General practitioner

CDM

Clinical decision-making

FG

Focus group

Authors’ contributions

Conceptualization: YC, LJ, LW and ZX; Data curation: YC, LJ, and WG; Formal analysis: YC and ZX; Investigation: LW, WG, YT, YG and ZX; Methodology: JD, LJ, YT, MY and XY; Funding acquisition: LW and ZX; Resources: LW and ZX; Supervision: ZX; Validation: YC, JW, JZ and ZX; Writing - original draft: YC and LJ; Writing - review & editing: all authors.

Funding

This study was supported by the Scientific Research Fund of the Zhejiang Provincial Education Department (Grant no. Y202352176), Educational Reform Project of Zhejiang University School of Medicine (Grant no. jgyb2025051), and 2025 Teaching Reform Research Project of the Second Clinical Medical College of Zhejiang University (Grant no. 20250104).

Data availability

The data generated or analyzed during this study can be provided by the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of the Second Affiliated Hospital of Zhejiang University School of Medicine [Approval No: (2024) Ethics Review for Research No. (1048)] and was conducted in accordance with the principles outlined in the Declaration of Helsinki. Written informed consent was obtained from GP residents involved in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.