Construction of a faculty competency model for medical simulation education integrated with GenAI: a multi-method quantitative study based on the perspective of medical students

Abstract

Background

The rapid advancement of Generative Artificial Intelligence (GenAI) is reshaping the landscape of medical simulation education, necessitating the enhancement of faculty competencies to effectively integrate evolving knowledge systems with GenAI for collaborative decision-making. However, current educational technologies face systemic limitations, including fragmented functionality, a disconnect between conventional and GenAI-driven teaching approaches, and a lack of dynamic capability assessment tools. Constructing a standardised capability scale is crucial to overcoming adaptation bottlenecks.

Methods

Grounded in the integrated Technological Pedagogical Content Knowledge (TPACK) framework, this research proposed a two-round Delphi method to construct a faculty competency assessment scale, with stratified sampling involving 434 participants from clinical medicine, medical technology, and nursing fields. Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) were used to test model fit (CFI = 0.939, RMSEA = 0.117). This proposed research was not registered as a clinical trial.

Results

The resulting 16-item scale exhibited strong psychometric properties, demonstrating excellent internal consistency (Cronbach’s α = 0.979), sampling adequacy (KMO = 0.961), and significant sphericity (Bartlett’s test, p < 0.001). No factor covariance was detected, and item consensus was high (Kendall’s W = 0.761, p < 0.001). The model showed acceptable fit (χ²/df = 6.893).

Conclusion

This empirically validated and standardised assessment model provides robust support for integrating GenAI with simulation-based health sciences education. It offers a foundational framework for advancing faculty competency development and fostering adaptive, technology-enhanced teaching practices, while noting that the scale primarily captures ethical awareness rather than comprehensive governance competence.

Trial Registration

Not applicable. As this research did not involve clinical interventions, trial registration was not required.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12909-026-09734-2.

Background

Generative Artificial Intelligence (GenAI) is fundamentally reshaping the landscape of medical simulation education by dynamically integrating clinical guidelines, evidence-based resources, and multi-modal case data into a continuously evolving knowledge network [1]. This transformation is not merely technical; it redefines pedagogical roles, fostering a triadic model of collaborative decision-making among students, AI, and educators [2]. As a result, simulation educators must transition from traditional “knowledge authorities” to “knowledge navigators,” requiring advanced competencies in critically screening, validating, and adapting AI-generated content throughout the technology adoption curve (honeymoon, frustration, adaptation, and acceptance phases).

The existing studies primarily explore GenAI through fragmented lenses such as isolated applications (e.g. virtual case generation) or basic teacher training workshops (like AI tool workshops)—without addressing the need for systemic integration of GenAI into the broader framework of simulation-based health sciences education [3]. Two major gaps persist: first, a disproportionate emphasis on technical functionalities at the expense of holistic pedagogical transformation; second, the absence of dynamic, learner-centred analyses that reflect the evolving cognitive needs of students and the resulting generational shift in educational expectations [4]. This misalignment hampers the effectiveness of current faculty development strategies in supporting sustainable, AI-driven transformation.

To address these challenges, this study describes the construction of a standardised faculty competency model tailored to GenAI-integrated medical simulation education. By deconstructing the synergistic relationship between traditional teaching competencies and GenAI-specific capabilities [4], the model provides a validated empirical foundation to inform faculty development and institutional reform in the intelligent era. Rather than proposing a new theoretical framework beyond TPACK, this study operationalises and contextualises TPACK competencies for GenAI-enhanced medical simulation education, where epistemic uncertainty, content volatility, and ethical risk are substantially amplified.

Methods

Aim

The main aim of this research is to construct and validate a competency model integrating GenAI with traditional instructional competencies for medical simulation educators.

Study design

A multi-method approach was employed incorporating the Delphi technique [5] and a cross-sectional survey for empirical validation. Details of the survey questionnaire are provided (see Questionnaire SDC 1, which provides the full survey instrument).

All procedures in both the Delphi process and the cross-sectional survey adhered to established ethical principles for survey-based educational research. Before participation, all experts and learners received clear information regarding the study purpose, procedures, data use, and their right to withdraw at any stage without any consequences. No identifiable personal information was collected during expert consultation, pilot testing, or scale validation. Questionnaire responses were anonymised at the point of data collection and stored securely with restricted access to the research team. As the study involved minimal risk and no clinical intervention, the methodological procedures met standards for low-risk educational research.

Scale development and validation

Technological Pedagogical Content Knowledge (TPACK)

The competency assessment scale was developed based on the Technological Pedagogical Content Knowledge (TPACK) framework, which emphasises the dynamic interaction among content knowledge (CK), pedagogical knowledge (PK), and technological knowledge, serving as a foundational model for faculty competency development in the artificial intelligence era [6].

The scale included two primary dimensions: (1) Traditional Teaching Competence (9 items), grounded in the integration of CK and PK, focused on core pedagogical abilities in medical simulation education. Items such as “Clinical Case Teaching Design Ability” reflected Shulman’s concept of pedagogical content knowledge, emphasizing the alignment of teaching strategies with disciplinary content [7]. (2) GenAI Integration Ability (7 items), designed in accordance with TPACK’s principle of adaptation of contextual technology, assessed educators’ capacity to align GenAI tools with specific simulation scenarios. Items such as “Dynamic Matching of Technical Tools and Simulation Scenarios” and “Information Reliability Assessment” addressed ethical awareness and instructional alignment, ensuring that technology use enhances rather than overshadows pedagogy, and avoiding the risk of tool alienation.

The item development adhered to TPACK’s context-sensitive and ethically aware principles, incorporating elements such as interprofessional collaboration and information credibility to address challenges specific to GenAI integration in health sciences training.

Ethical awareness were indirectly assessed through items related to information reliability, safe use of GenAI tools, and responsible guidance, rather than through direct measures of ethical reasoning or governance competence.

Delphi method design and verification

To ensure theoretical rigour and practical relevance, experts were required to meet both of the following criteria: (1) at least 10 years of clinical teaching experience, and (2) demonstrated familiarity with the TPACK framework.

Familiarity with the TPACK framework was verified through an eligibility form requiring experts to document prior TPACK-related teaching, training, or publication experience, which was additionally cross-checked through institutional teaching records.

In the first round of the Delphi process, five experts participated, including three clinical care specialists and two health sciences education scholars. All held senior professional titles and had extensive experience in simulation-based teaching and faculty development. Although the number of experts in the first Delphi round was relatively small (n = 5), which may limit the breadth of disciplinary perspectives and the stability of early consensus, rigorous inclusion criteria and high agreement indices were applied to mitigate this limitation. To ensure balanced representation of the “technology–pedagogy–content” triad emphasised in the TPACK model, the content validity standard proposed by Beck (S-CVI/Ave ≥ 0.90) was employed [8]. This rigorous threshold helped prevent undue emphasis on technical functionality during item formulation and promoted alignment with educational theory.

Following distribution of the initial questionnaire, expert agreement was evaluated using Kendall’s coefficient of concordance, revealing a high level of consensus (W = 0.83, χ² = 53.12, df = 4, p < 0.001). Based on qualitative feedback, several items were revised for improved clarity and operational precision—for example, clearer definitions were provided for “teamwork competence” and “GenAI information discernment ability.” In the second round of consultation, the revised scale demonstrated even stronger expert consensus (W = 0.84, χ² = 53.50, df = 4, p < 0.001), indicating increasing convergence toward a theoretically coherent and practically relevant instrument.

To further verify content validity, a validation panel of 10 participants was assembled, including five experts in medical simulation education (three clinical educators and two health sciences education researchers, all with senior professional titles and over 10 years of experience) and five learners (three senior undergraduate interns and two resident physicians), each with exposure to more than 20 simulation-based training sessions. Using a four-point Likert scale (1 = not relevant, 4 = highly relevant), item-level content validity indices (I-CVI) ranged from 0.70 to 1.00. At the scale level, both the universal agreement index (S-CVI/UA = 0.81) and the average agreement index (S-CVI/Ave = 0.96) met the quality benchmarks recommended by Polit and Beck [8], confirming the strong content validity of the scale, indicating that the revised items more accurately reflected both technological integration and pedagogical expectations in medical simulation contexts.

In this study, expert and learner content validity index (CVI) ratings complementary, but not interchangeable roles. Expert CVI ratings were primarily used to evaluate the theoretical relevance, conceptual representativeness, and domain coverage of the proposed competency items, consistent with the original intent of the Delphi technique. Learner CVI ratings were incorporated to assess item clarity, interpretability, and contextual relevance from the end-user perspective in simulation-based education settings. The combination of expert and learner CVI ratings was therefore intended to triangulate content validity and enhance ecological validity, rather than to replace expert judgment.

Preliminary experiment design and verification

Respondents were randomly selected from individuals who had completed at least ten sessions of medical simulation-based training. To ensure immediacy and contextual relevance, questionnaires were distributed within 30 min after the course. All 30 responses were valid, yielding a 100% effective response rate. Sample characteristics are detailed in Table 1.

Table 1.

Characteristics of pre-test sample (n = 30)

‌Item	‌Content‌	‌Frequency‌	‌Percentage (%)‌	‌Cumulative Percentage (%)‌
‌Gender‌	Male	20	66.67	66.67
	Female	10	33.33	100.00
‌Age Group‌	18–25 years	28	93.33	93.33
	31–40 years	2	6.67	100.00
‌Education Level‌	Associate Degree (In Progress)	1	3.33	3.33
	Bachelor’s Degree	10	33.33	36.67
	Master’s Degree	19	63.33	100.00
‌Major Category	Clinical Medicine	25	83.30	83.30
	Medical Technology	3	10.00	93.30
	Nursing	2	6.67	100.00
‌GenAI Application Experience‌	Never used	9	30.00	30.00
	Used 1–2 types	15	50.00	80.00
	Used 3–5 types	6	20.00	100.00

The assessment instrument employed a five-point Likert scale (1 = very unimportant, 5 = very important), with total scores ranging from 16 to 80. Based on principal component analysis, two latent factors were extracted, jointly explaining 69.086% of the total variance (Fig. 1). The first factor, representing traditional medical simulation teaching competency, accounted for 38.674% of the variance and featured high loadings on items such as “Communication Skills” (loading = 0.862) and “Course Design and Organisation Skills” (loading = 0.859). The second factor, reflecting the integration and innovation of GenAI technology in medical simulation education, accounted for 30.412% of the variance, with strong item loadings such as “GenAI Ethics and Safety Knowledge” (loading = 0.863) and “Ability to Guide Students in Using GenAI for Learning” (loading = 0.818).

Fig. 1.

Scree plot of the principal component analysis showing the eigenvalues of extracted factors

Content validity was established through reference to theoretical framework for integrated health sciences education, ensuring comprehensive coverage of both pedagogical and technological domains. The overall internal consistency of the scale was excellent, with a Cronbach’s α coefficient of 0.932. Reliability for each dimension was also high, with α = 0.948 for the traditional teaching competency domain and α = 0.900 for the GenAI integration domain. The “GenAI integrated medical simulation education faculty competency needs table” was shown in Table 2.

Table 2.

Demand Scale for Educator Competency in GenAI-Integrated Medical Simulation Education (16–80 points)

‌Category‌	‌Item‌	‌1‌	‌2‌	‌3‌	‌4‌	‌5‌
‌Traditional Medical Simulation Teaching & Literacy Competency‌	‌1. Solid Medical Expertise‌: Mastery of foundational and clinical medical knowledge; ability to accurately explain diseases.
	‌2. Rich Clinical Experience‌: Extensive clinical practice and case integration for teaching.
	‌3. Effective Teaching Methodology‌: Proficiency in diverse teaching methods and adaptive application.
	‌4. Strong Communication Skills‌: Clear and precise communication with students and patients, providing actionable guidance.
	‌5. Learning Facilitation‌: Guiding students to develop learning plans and autonomous study habits.
	‌6. Curriculum Design & Organization‌: Logical structuring of content and progress.
	‌7. Teaching Evaluation‌: Objective assessment of student outcomes and feedback delivery.
	‌8. Team Collaboration‌: Effective coordination with educators and healthcare professionals to foster student development.
	‌9. Professional Ethics & Humanistic Literacy‌: Adherence to medical ethics, patient-centered care, and exemplary conduct.
GenAI-Technology Integration & Innovation Competency‌	‌10. GenAI Tool Proficiency‌: Skilled use of generative AI tools (e.g., ChatGPT, DeepSeek, DALL-E, medical diagnostic AI) for lesson preparation and case generation.
	‌11. GenAI Content Verification‌: Ability to evaluate accuracy and reliability of AI-generated medical information.
	‌12. GenAI-Driven Curriculum Design‌: Leveraging AI to analyze student data and tailor personalized teaching plans.
	‌13. GenAI Ethics & Safety‌: Knowledge of ethical guidelines (e.g., data privacy, AI diagnostic accountability) for medical AI applications.
	‌14. GenAI-Guided Learning‌: Training students to use GenAI tools for knowledge acquisition and case analysis with critical thinking.
	‌15. GenAI Data Analysis‌: Interpreting AI-generated teaching data (e.g., student knowledge/skill metrics) to evaluate outcomes.
	‌16. GenAI-Teaching Integration Innovation‌: Pioneering novel methods (e.g., AI-enhanced simulation cases) to diversify pedagogy.

Sampling and evaluation

In this study, the term “learners” is used as an umbrella term referring collectively to senior undergraduate interns and resident physicians who participated in simulation-based training. Sample size estimation using G*Power 3.1 software determined that a minimum of 418 participants were required for adequate statistical power (see Table SDC 2, which provides analysis results of power calculation; Table SDC 3, which provides Input parameters used for power analysis). Participants were drawn from professional healthcare personnel who had attended at least ten medical simulation sessions between February 14 and March 28, 2025, at a tertiary hospital’s Clinical Skills Training Center (Ethical Approval No. XY-2024-003). All data were de-identified before analysis.

Proportional stratified sampling was implemented based on historical training participation rates across three professional categories: clinical medicine, nursing, and medical technology (approximate ratio 6:1.5:1.3). Attendance records (N = 1286) served as the sampling frame, including job type (e.g., interns or learners within one-year post-graduation). Within each stratum, subjects were assigned unique identification codes. Random sampling was conducted using Excel’s RANDBETWEEN function, with duplicates removed and additional draws performed as needed to meet target quotas (see Table SDC 4, which offers results of random sampling by stratum).

Data analysis

Descriptive statistics were used to summarise demographic characteristics and item responses. Internal consistency was evaluated using Cronbach’s α coefficient [9]. Factor structure was explored using principal component analysis (PCA), followed by Confirmatory Factor Analysis (CFA) to assess structural validity and model fit [10]. Kendall’s W test was used to evaluate expert consensus during Delphi rounds, ensuring consistency in item development and refinement [11]. These analyses supported the final refinement of the scale and confirmed its alignment with the proposed theoretical model.

Results

Study design and participants

A total of 434 valid questionnaires (see Table SDC 5, which provides Baseline valid questionnaire). The distribution of “total scores” is shown in Table SDC 6, with a total score ranging from 28 to 80 points. The minimum value (M) is 28 points, the maximum value (X) is 80 points, and the average value (E) is 72.46 points, indicating an overall level slightly above average.

Reliability, validity, and factorial structure of the faculty competency scale

The internal consistency reliability of the scale was assessed using Cronbach’s alpha, yielding a coefficient of 0.979 across 16 items. This result indicates excellent internal consistency, demonstrating that the instrument possesses high reliability and stability in measuring faculty competencies in medical simulation education. To further explore the underlying structure of the competency variables, PCA was conducted. The Kaiser-Meyer-Olkin (KMO) measure was 0.961, suggesting strong sampling adequacy, while Bartlett’s test of sphericity was significant (χ² = 9881.023, df = 120, p < 0.001), confirming the suitability of the data for factor analysis. The component score covariance matrix indicated zero covariance between the two extracted components, demonstrating that the rotated factors were orthogonal and uncorrelated.

The total variance explained by the first two principal components was substantial, with eigenvalues of 12.174 and 1.241, accounting for 76.085% and 7.755% of the variance, respectively, yielding a cumulative variance of 83.840%. The scree plot illustrates a marked drop in eigenvalues after the second component, supporting the decision to retain two factors (Fig. 2). The unrotated component matrix revealed initial loadings of variables on the components, and after rotation (converged in three iterations), the rotated solution provided a clearer factor structure. The first component showed high loadings on traditional teaching competencies such as teamwork, curriculum design and organisation, and assessment of teaching effectiveness. The second component demonstrated high loadings on competencies related to GenAI, including the ability to guide students in using GenAI for learning and the ability to analyse data using GenAI-related tools.

Fig. 2.

Scree plot showing the eigenvalues of the principal components

Communality estimates showed that all variables were well represented by the extracted factors. For example, the communality for the ability to guide students in GenAI-based learning was 0.913, and for teamwork, it was 0.868, indicating that the two principal components effectively captured the underlying variance in these measures. Finally, a Kendall’s coefficient of concordance analysis based on the 16 items revealed a strong consensus among participants (W = 0.761, χ²(433) = 5270.561, p < 0.001), suggesting that medical learners exhibited a high degree of agreement in their ranking of faculty competency dimensions. In the participant survey, Kendall’s W was used as a complementary index to reflect the degree of shared perception among learners regarding the relative importance of faculty competency domains, rather than as a substitute for factor analytic methods.

Results of confirmatory factor analysis

Factor loadings and convergent validity results are presented in Table SDC 7. The latent construct “Traditional Competencies in Medical Simulation Education” was measured by nine observed variables, with standardised factor loadings ranging from 0.842 to 0.935, all statistically significant (p < 0.05). The construct “Innovative Integration of GenAI in Medical Simulation Education” was measured by seven observed variables, with standardised loadings ranging from 0.816 to 0.961, also all were significant (p < 0.05). All items demonstrated high factor loadings (> 0.80), indicating strong explanatory power for the corresponding latent constructs. The Average Variance Extracted (AVE) for the two constructs were 0.806 and 0.821, respectively, while their Composite Reliability (CR) values were 0.974 and 0.970. These results satisfy the commonly accepted thresholds of Average Variance Extracted (AVE) > 0.5 and Composite Reliability (CR) > 0.7, indicating strong internal consistency and good convergent validity. As shown in Table SDC 8, the square roots of AVE values exceeded the off-diagonal Pearson correlation coefficients, confirming satisfactory discriminant validity between the two latent constructs.

The results of model fit analysis are presented in Fig. 3. The chi-square to degrees of freedom ratio was χ²/df = 6.893, with CFI = 0.939, RMSEA = 0.117, RMR = 0.012, and NFI = 0.929. While some indices approached the ideal thresholds, the overall model demonstrated moderate model fit. The standardised covariance estimate between the two latent constructs was 0.837, with a z-value of 11.323 (p < 0.05), indicating a significant positive correlation. Regarding residuals, all measurement item residual estimates had z-values greater than 10 with p = 0, supporting the appropriateness of residual estimation and further confirming the model’s validity.

Fig. 3.

Model fit indices of the confirmatory factor analysis

It should be noted that RMSEA values may be inflated in models with high factor loadings and complex item structures, and therefore should be interpreted cautiously and in conjunction with other fit indices.

Discussion

Existing studies on technology-integrated medical education have highlighted the need for validated frameworks that capture the evolving intersection of pedagogy and emerging technologies. Building on this foundation, the present study provides several novel contributions.

Theoretical contributions

Rather than proposing a new theoretical framework beyond TPACK, the present study operationalises and contextualises TPACK competencies within GenAI-enhanced medical simulation education. The contribution of the Med-TPACK Scale lies in translating emerging GenAI-related pedagogical challenges—such as epistemic uncertainty, content volatility, and ethical risk awareness—into measurable faculty competencies. This study advances the theoretical understanding of faculty competency in GenAI-integrated medical simulation education by addressing the absence of a standardised assessment framework in this emerging field. Existing literature has highlighted the need to re-conceptualise teaching competence in technology-enhanced environments, particularly with the rapid adoption of AI-assisted tools in health professions education. Recent reviews have further emphasised that current AI-supported medical education lacks unified frameworks for evaluating educator readiness and digital capability [12, 13]. Building on the TPACK framework, our study extends prior work by incorporating new competency dimensions such as digital literacy, contextualised feedback, and AI-related ethical awareness, which have been increasingly emphasised in recent medical education research. By integrating these dimensions into a coherent structure, the Med-TPACK Scale contributes to an emerging next-generation competency paradigm for simulation-based teaching, which emphasises adaptability to rapidly changing technological ecosystems.

Moreover, the conceptual refinement of technologically enhanced content knowledge and ethical-constraint technological knowledge aligns with current debates regarding how educators should critically evaluate and responsibly integrate GenAI tools in simulation-based teaching. Recent scoping reviews have similarly stressed that responsible and transparent use of AI requires explicit measurement of ethical awareness competencies in clinical education settings [14]. These expansions provide a theoretical foundation for understanding how traditional simulation competencies and GenAI-driven instructional capabilities intersect and co-evolve. Rather than reiterating empirical indicators, the present study highlights how these constructs collectively advance a more comprehensive and future-oriented framework for faculty capability building.

Practical implications

The findings offer several practical implications for faculty development in simulation-based medical education. First, the Med-TPACK Scale provides a structured and measurable approach to identifying competency needs related to GenAI integration. Prior studies have emphasised that educators often adopt new technologies without sufficient pedagogical or ethical preparation; our scale helps address this gap by enabling institutions to diagnose specific areas requiring support. This aligns with calls in recent research suggesting that AI adoption in medical education frequently outpaces faculty preparedness [15].

Second, by capturing competencies across both traditional and GenAI-enhanced domains, the scale can inform precision-oriented professional development programs. Institutions may use the tool to design tailored training pathways that align educator capability with instructional demands in mixed-reality, data-driven, or AI-augmented simulation environments. This supports a shift from experience-based faculty training to systematic, data-informed enhancement strategies.

Finally, the scale has potential applications in curriculum planning and policy development. As simulation education increasingly incorporates AI-enabled tools, policymakers may use the competencies identified in this study to establish clearer expectations for faculty readiness and to guide resource allocation for digital transformation in medical education.

Limitations and future directions

This study has several limitations that should be considered when interpreting the findings. First, the relatively small number of experts in the initial Delphi round, together with the regionally and culturally homogeneous study sample, may constrain the robustness and generalisability of content validity judgments, despite high consensus indices. While rigorous inclusion criteria and agreement measures were applied, future studies should involve larger, more diverse, and cross-institutional expert panels, including specialists in areas such as AI algorithm design and healthcare data science, to enhance external validity.

Second, both PCA and CFA were conducted on the same dataset, which may increase the risk of overfitting and optimistic model fit estimates. Independent sample validation is therefore warranted in future research.

Third, although the Med-TPACK Scale demonstrated excellent internal consistency, the extremely high Cronbach’s α coefficient (> 0.97) may also indicate potential item redundancy, suggesting that some items capture highly overlapping aspects of faculty competence rather than distinct constructs. As noted in prior psychometric literature, excessively high reliability may reflect redundancy rather than optimal measurement precision. Future studies may consider item reduction strategies, alternative reliability indices such as McDonald’s ω, or bifactor modelling approaches to further refine the scale structure.

Finally, while the scale shows strong theoretical grounding and statistical validity, its performance has not yet been examined in authentic clinical or high-fidelity simulation environments. Application-based validation across diverse instructional modalities, including VR, MR, and GenAI-augmented systems, is needed to assess its robustness and translational value in real-world teaching contexts.

Conclusion

Traditional medical simulation and GenAI technologies complement each other in health sciences education. Our findings provide empirical evidence for their synergistic integration, with key competency dimensions—such as clinical practice experience and GenAI-driven instructional optimisation—demonstrating substantial co-development potential. Advancing this integration requires comprehensive technology–pedagogy fusion frameworks, supported by interdisciplinary faculty training, competency-based interventions, and continuous optimisation of the instructional ecosystem. Moreover, ethical and governance considerations of GenAI remain crucial; however, conclusions should be interpreted cautiously, as the current scale captures ethical awareness, rather than full ethical reasoning or regulatory competence.

Supplementary Information

Supplementary Material 1. ^{(53.5KB, doc)}

Supplementary Material 2. ^{(14.5KB, doc)}

Supplementary Material 3. ^{(17KB, doc)}

Supplementary Material 4. ^{(16KB, doc)}

Supplementary Material 5. ^{(29.5KB, doc)}

Supplementary Material 6. ^{(12KB, doc)}

Supplementary Material 7. ^{(14KB, doc)}

Supplementary Material 8. ^{(13KB, doc)}

Supplementary Material 9. ^{(979.6KB, pdf)}

Supplementary Material 10. ^{(24KB, docx)}

Abbreviations

GenAI

Generative Artificial Intelligence

TPACK

Technological Pedagogical Content Knowledge

EFA

Exploratory Factor Analysis

CFA

Confirmatory Factor Analysis

CK

Content knowledge

PK

Pedagogical knowledge

PCA

Principal component analysis

KMO

Kaiser-Meyer-Olkin

AVE

Average Variance Extracted

CR

Composite Reliability

CFI

Comparative Fit Index

RMSEA

Root Mean Square Error of Approximation

NFI

Normed Fit Index

SEM

Structural Equation Modelling

S-CVI/UA

Scale-level Content Validity Index based on Universal Agreement

Authors’ contributions

Conceptualisation: YZG (lead), HXY (supporting); Data curation: XTY; Formal analysis: YZG (lead), XTY (supporting); Funding acquisition: HXY; Investigation: XTY；Methodology: YZG (lead), XTY(supporting); Project administration: YZG (lead), XTY (supporting); Resources: QC；Supervision: HXY; Validation: HXM (lead), YZG (supporting); Visualisation: HXM; Writing – original draft: YZG (lead), XTY (supporting); Writing – review & editing: XTY(lead), YZG (supporting).

Funding

This study was supported by the Special Fund for Clinical Scientific Research of Wu Jieping Medical Foundation (Grant No. 3206750, 2020-10-36).

Data availability

We anonymised data which is currently stored in our local data warehouse. The datasets used and analysed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Ethics Committee of Wenzhou Medical University Second Affiliated Hospital (Approval No. 2025-K-26-01, Supplementary File SDC9). Both the Delphi and cross-sectional surveys were conducted in accordance with ethical principles. Data collection was carried out between February 14 and March 28, 2025. All participants were fully informed about the study objectives and procedures. Informed consent was implied through voluntary completion of the questionnaire, as stated in the questionnaire introduction (Supplemental Digital Content 10). Participation was entirely voluntary. Measures were implemented to ensure anonymity, confidentiality, and data integrity, thereby protecting participants’ rights and promoting responsible conduct of research. As this study did not involve clinical interventions, trial registration was not required.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.