Preliminary evaluation of an AI-enhanced simulated interprofessional learning intervention for pharmacy students: a mixed-methods pilot study Authors: Iqbal Fahs, Mariam Dabbous, Jihan Safwan, Seham Kanaan, Faten Hamed, Mariam Alkhatib, Jana Hayek, Faten Ezzeddine, Haya Hamed, Fouad Sakr, Mohamad Rahal Published: 2026 ABSTRACT BackgroundInterprofessional education (IPE) is vital for preparing pharmacy students for collaborative practice but is often constrained by logistical challenges. This study examined the preliminary impact of an AI-enhanced simulated interprofessional learning intervention on pharmacy students’ self-reported interprofessional competencies and perceived readiness for AI use in healthcare, using an antimicrobial stewardship (AMS) scenario to assess the application of clinical decision-making.MethodsA pilot quasi-experimental study with an explanatory mixed-methods design was conducted among pharmacy students (n = 112). The intervention involved an AI-enhanced simulation using virtual avatars representing healthcare roles within a recurrent urinary tract infection case. Outcomes were assessed using the Interprofessional Collaborative Competency Attainment Survey (ICCAS), the Medical Artificial Intelligence Readiness Scale (MAIRS), and an AMS knowledge questionnaire. Paired t-tests, repeated-measures MANCOVA, and multiple linear regression analyses were performed, and thematic analysis was used for qualitative data.ResultsParticipants reported significantly higher post-intervention self-reported scores across all outcomes. ICCAS scores increased from 58.82 ± 14.67 to 68.52 ± 12.45, MAIRS from 75.16 ± 10.85 to 82.12 ± 9.49, and AMS from 6.21 ± 1.74 to 8.08 ± 1.35 (all p < 0.001), with all subscales showing significantly higher post-values (p < 0.001). MANCOVA confirmed a significant overall time effect (p < 0.001). Qualitative findings highlighted students’ perceived improvements in communication, teamwork, clinical reasoning, and critical awareness of AI limitations. Higher academic year, prior IPE exposure, simulation experience, and digital literacy were significantly associated with higher outcome scores. Students without prior IPE exposure showed larger pre-post ICCAS change scores than those with prior IPE exposure (Δ difference = 4.2, p = 0.008, d = 0.52).ConclusionThis pilot study suggests that AI-enhanced simulated interprofessional learning may be associated with improvements in students’ self-reported collaborative competencies, perceived readiness for AI use in healthcare, and short-term AMS knowledge scores. Given the single-group pre-post design and short-term assessment, findings should be interpreted as preliminary and require confirmation through future controlled randomized and longitudinal studies.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12909-026-09643-4. Introduction The increasing complexity of healthcare delivery necessitates effective collaboration among healthcare professionals [1]. Interprofessional collaboration (IPC) has been consistently associated with improved clinical outcomes, reduced medical errors, and enhanced patient safety, particularly in areas requiring coordinated pharmacotherapy decisions and shared clinical decision-making [2, 3]. As pharmacists assume increasingly patient-centered roles within multidisciplinary teams, the ability to communicate effectively, clarify professional responsibilities, and contribute to collaborative care planning has become a core competency in pharmacy practice [4]. Interprofessional education (IPE) is widely recognized as a foundational strategy for preparing healthcare students for collaborative practice [5]. Frameworks developed by organizations such as the Interprofessional Education Collaborative (IPEC) and the World Health Organization (WHO) emphasize the integration of IPE into health professions curricula [6–8]. Evidence demonstrates that IPE can improve pharmacy students’ communication skills, clinical reasoning, and the ability to engage in team-based decision-making [5, 9]. However, despite its recognized importance, the implementation of IPE remains challenging. Logistical barriers including scheduling conflicts across programs, limited faculty resources, and infrastructural constraints often restrict the scalability and consistency of interprofessional learning experiences [10, 11]. These challenges are particularly pronounced in institutions with multiple campuses and large student cohorts, highlighting the need for innovative and flexible educational approaches. Simulation-based education has emerged as an effective pedagogical strategy to address these limitations by providing structured and safe learning environments in which students can develop clinical and collaborative competencies [12, 13]. Within pharmacy education, simulation is increasingly used to enhance clinical decision-making, patient counseling, and medication management skills. More recently, advances in artificial intelligence (AI) have expanded the potential of simulation by enabling dynamic, adaptive, and interactive learning experiences [14]. AI-powered tools can simulate real-time interactions with patients and healthcare professionals, allowing learners to engage in role-based communication, iterative clinical reasoning, and context-specific decision-making [15–17]. Importantly, AI-enhanced simulation offers the potential to replicate key elements of interprofessional interaction without requiring the physical co-presence of multiple disciplines, thereby addressing a major limitation of traditional IPE models. Despite this potential, the integration of AI into interprofessional learning remains limited. Existing research has primarily focused on AI applications in individual learning or clinical decision support, with comparatively little attention to its role in facilitating interprofessional competency development [18]. In addition, readiness to engage with AI technologies has emerged as an essential outcome in health professions education, reflecting the growing importance of digital literacy, critical appraisal of AI outputs, and ethical use of technology in clinical practice [19]. However, these competencies are not yet consistently embedded within interprofessional educational frameworks [20]. Together, these gaps highlight the need for innovative models that integrate AI, simulation, and interprofessional learning within a unified, competency-based approach tailored to contemporary pharmacy practice. To operationalize such a model, the selection of an appropriate clinical context is essential. Educational scenarios that require high levels of interprofessional coordination, structured clinical reasoning, and evidence-based decision-making are particularly well suited for simulated interprofessional learning. Antimicrobial stewardship (AMS) represents one such context. The optimization of antimicrobial therapy involves complex decision-making processes, including appropriate drug selection, dosing, duration, and de-escalation, and requires close collaboration among physicians, pharmacists, nurses, and other healthcare professionals [21, 22]. Pharmacists play a central role in AMS programs, contributing to guideline implementation, therapeutic optimization, and prevention of antimicrobial resistance [23]. As such, AMS provides a clinically relevant and pedagogically appropriate framework through which interprofessional competencies and clinical reasoning can be simultaneously developed and assessed. At the Lebanese International University School of Pharmacy, efforts to integrate interprofessional education into the curriculum have previously demonstrated positive effects on students’ collaborative competencies and perceptions [24]. However, the implementation of traditional IPE remains constrained by challenges such as coordination across programs, variability in clinical training environments, and the need to deliver education across multiple campuses and large student cohorts. These factors limit consistent access to structured interprofessional experiences for all pharmacy students. Hence, there is a need to explore innovative approaches that can sustain and expand interprofessional learning opportunities while addressing these logistical constraints. Given the novelty of integrating AI-enhanced virtual role interaction within simulated interprofessional learning, and the limited existing evidence regarding its implementation in pharmacy education, this study was designed as a preliminary pilot evaluation intended to explore feasibility, acceptability, and potential educational impact prior to future controlled investigations. Therefore, this pilot study aimed to evaluate the preliminary impact of an AI-enhanced simulated interprofessional learning intervention on pharmacy students’ self-reported interprofessional collaboration competencies and perceived readiness for AI use in healthcare. To provide a clinically meaningful and interprofessionally intensive learning context, the intervention was anchored in an antimicrobial stewardship scenario involving the management of recurrent urinary tract infection (UTI). A planned subgroup analysis further explored whether prior IPE exposure may influence pre-post score differences associated with the intervention. Although grounded in IPE principles, the intervention involved pharmacy students interacting with AI-powered avatars rather than learners from multiple disciplines, representing simulated interprofessional learning rather than true IPE. Design and methods Study design and sampling This pilot study employed a quasi-experimental pre-post intervention design to evaluate a structured simulated interprofessional learning intervention enhanced by AI-supported virtual role interaction. The study adopted an explanatory sequential mixed-methods design, in which quantitative data were collected and analyzed first, followed by qualitative data collection to provide deeper insight into participants’ experiences and the observed findings. Integration of quantitative and qualitative findings was performed during the interpretation stage to achieve a more comprehensive understanding of the intervention outcomes. Given the exploratory nature of this novel educational approach, the study was designed as a preliminary pilot investigation intended to generate foundational evidence and inform future controlled studies. The intervention was implemented during the Fall semester of the academic year 2025–2026. All eligible third-, fourth-, and fifth-year pharmacy students enrolled at the Lebanese International University School of Pharmacy during the study period were invited to participate. A total of 559 eligible students were invited, including 206 third-year students, 189 fourth-year students, and 164 fifth-year students. Of these, 112 students agreed to participate and completed both the pre- and post-intervention assessments, corresponding to a response rate of 20.0% and a completion rate of 100% among participating students. Participation was voluntary, and only students with complete paired responses were included in the final analysis. A pragmatic census-based recruitment approach was used whereby all eligible students exposed to the educational activity were invited to participate in the study. As this was an exploratory pilot study using a census-based recruitment strategy, no a priori sample size calculation was performed. Surveys were distributed electronically via institutional email one week before the intervention (pre-assessment) and one week after the intervention (post-assessment). Pre- and post-intervention responses were matched using anonymous identification codes. Ethical approval This study was conducted in accordance with the Ethical Guidelines for Life Science and Medical Research Involving Human Subjects and the principles of the Declaration of Helsinki. Ethical approval was obtained from the Ethical and Research Committee of the Lebanese International University (Approval No. 2025ERC-021-LIUSOP). Participation was entirely voluntary, and informed consent was obtained from all participants prior to access to the study questionnaires. Students were informed that participation and responses would have no impact on their academic evaluation. Participant anonymity and data confidentiality were strictly maintained throughout the study. As this investigation did not involve a clinical intervention, clinical trial registration was not applicable. Intervention The AI-enhanced simulated interprofessional learning intervention was informed by a formal needs assessment conducted within the School of Pharmacy at the Lebanese International University. The intervention was intentionally mapped to established competency frameworks and guidance encompassing interprofessional collaboration, responsible use of AI, and AMS [7, 19, 25]. It was sequenced through foundational didactic sessions, AI-enhanced simulated interprofessional learning activity, team-based clinical reasoning, and facilitated debriefing to ensure comprehensive coverage of these competency domains. The intervention was delivered over a structured one-day session lasting approximately six hours. The detailed schedule, session objectives, and competency mapping across IPEC [7], WHO guidance on ethics and governance of AI for health [19], and WHO competency framework for health workers’ education and training on AMR [25] are presented in Supplementary Table 1. A schematic overview of the four-phase simulation workflow is presented in Fig. 1. Before the simulation, participants attended three orientation sessions addressing core IPE principles, AMS, and responsible use of AI in healthcare. These sessions provided a shared conceptual foundation and prepared students for effective engagement in the simulated activity. Subsequently, students participated in a case-based AI-enhanced simulated interprofessional activity centered on the management of a recurrent UTI. Participants were assigned to teams of 7 to 10 students from the three different professional pharmacy years. The simulation incorporated AI-powered virtual avatars representing key healthcare roles, including a patient, physician, nurse, and dietitian. These avatars were intentionally designed to simulate realistic interprofessional dynamics, such as time constraints, limited availability, and profession-specific priorities. Students were required to actively engage with each role, clarify responsibilities, and integrate information from multiple professional perspectives to inform their clinical decisions. Students were instructed to rely on evidence-based clinical guidelines, AMS principles, and interprofessional consensus when developing management plans, rather than deferring decision-making to AI-generated responses only. Fig. 1Workflow of the AI-Enhanced Simulated Interprofessional Learning Intervention. The intervention consisted of four sequential phases including orientation sessions, AI-enhanced simulated interprofessional learning activity using AI-powered healthcare avatars and structured worksheets, collaborative care plan development with team presentations, and facilitated debriefing with reflective discussion. The activity integrated interprofessional education principles, responsible AI use, and antimicrobial stewardship concepts within a recurrent urinary tract infection simulation scenario Workflow of the AI-Enhanced Simulated Interprofessional Learning Intervention. The intervention consisted of four sequential phases including orientation sessions, AI-enhanced simulated interprofessional learning activity using AI-powered healthcare avatars and structured worksheets, collaborative care plan development with team presentations, and facilitated debriefing with reflective discussion. The activity integrated interprofessional education principles, responsible AI use, and antimicrobial stewardship concepts within a recurrent urinary tract infection simulation scenario The simulation followed a standardized workflow supported by a structured student worksheet. Initially, teams conducted focused history-taking with the patient avatar to identify presenting symptoms, recurrence patterns, medication adherence behaviors, prior antibiotic exposure, and relevant clinical risk factors. Subsequently, teams engaged in information gathering by interacting with physician, nurse, and dietitian avatars to obtain complementary clinical, nursing, and lifestyle-related insights, while practicing concise, profession-appropriate communication and role clarification. Teams then collaboratively developed and presented their management care plan and engaged in facilitated reflection guided by faculty members, focusing on interprofessional collaboration, communication challenges, decision-making processes, and ethical considerations encountered during the simulation. A total of 10 faculty facilitators participated in the activity, maintaining a facilitator-to-student ratio of approximately 1:10. To promote consistency and quality across small-group interactions, all facilitators underwent preparatory training prior to the activity. This preparation consisted of a two-hour orientation session outlining the intervention objectives, targeted competencies, and recommended facilitation approaches. Facilitators were also provided with a standardized guide detailing session objectives, case summaries, discussion prompts, suggested timelines, and techniques to encourage equitable participation among students from different academic years. Technical specification of the AI simulation platform The AI-powered avatars were implemented using ConvAI, a conversational AI platform designed for educational simulations [26]. Representative screenshots of the AI-enhanced simulation environment, guided starter prompts, and sample interactive student-avatar dialogues are provided in Supplementary Figs. 1–4 to improve transparency and reproducibility of the intervention. ConvAI utilizes large language models (LLMs) with role-specific system prompts to generate contextually appropriate responses from virtual characters. For this study, four avatar personas were created: a patient, physician, nurse, and dietitian. Each avatar was configured with a unique system prompt defining its professional role, communication style, clinical knowledge boundaries, and personality traits aligned with the recurrent UTI case scenario. The avatars operated in a semi-autonomous conversational mode, generating responses in real-time based on student questions and prompts. A faculty facilitator monitored each simulation session and had the ability to pause the simulation or provide verbal guidance if technical issues arose or if the AI generated clinically inappropriate content. To ensure consistency across all student teams, several standardization measures were implemented. First, each avatar’s system prompt was identical across all sessions, defining the same clinical knowledge base, personality, and response parameters. Second, all avatars were provided with the same case background, including patient history, previous antibiotic exposures, allergy status, and clinical findings. Third, the LLM temperature parameter was set to 0.3 (on a scale of 0 to 1) to balance response variability with clinical accuracy, reducing the likelihood of divergent or hallucinated responses. Fourth, facilitators followed a standardized script for introducing the simulation, answering student questions, and managing the workflow to ensure consistent delivery across teams. Although these standardization measures were implemented, minor variability in AI-generated responses between sessions may still have occurred due to the probabilistic nature of large language model outputs. Unlike traditional scripted virtual simulation activities that rely on predetermined dialogue trees or fixed response pathways, the AI-enhanced simulation used in this study allowed dynamic adaptive real-time interaction between students and avatars. Avatar responses dynamically evolved according to students’ questions, clinical reasoning processes, and conversational flow, enabling iterative clarification, role-based communication, and flexible interprofessional discussion. This adaptive interaction was intended to simulate more realistic clinical communication and collaborative decision-making processes. Data collection and measures Data were collected using a structured, self-administered questionnaire that was developed and administered in English. The questionnaire comprised both previously validated instruments and investigator-developed items. Interprofessional collaboration competencies were assessed using the Interprofessional Collaborative Competency Attainment Survey (ICCAS) [27, 28], and readiness for AI use in healthcare was assessed using the Medical Artificial Intelligence Readiness Scale (MAIRS) [29]. The AMS knowledge component was developed by the research team based on international AMS recommendations and aligned with the learning objectives of the intervention [25, 30]. The full questionnaire is provided in Supplementary File 1. The questionnaire was reviewed by three faculty members to assess content validity, clarity, and relevance to the study objectives. The instrument was subsequently pilot tested with a small group of students (n = 5), whose responses were excluded from the final analysis. Feedback from the pilot testing informed minor refinements, resulting in the finalized version of the survey used for data collection. The questionnaire comprised four main sections including sociodemographic and educational characteristics, interprofessional collaboration, readiness for AI use in healthcare, and AMS knowledge, in addition to open-ended reflective questions included in the post-intervention survey. Sociodemographic and educational variables included age, gender, cumulative grade point average (GPA), academic year (Third, fourth, or fifth year), prior training in AMS (Previous formal clinical training and application addressing appropriate antimicrobial use), previous participation in IPE activities (any prior structured learning experience involving two or more health disciplines aimed at promoting collaboration, role understanding, or team-based care), prior exposure to clinical simulation (Participation in any simulation-based educational activity designed to replicate clinical scenarios), attendance of AI-related courses or workshops, and prior use of AI tools in learning or clinical reasoning (Experience using AI-enhanced tools to support academic learning or clinical decision-making). Baseline perceptions of familiarity with AI in healthcare, perceived usefulness of AI for learning, and perceived digital literacy were assessed using 5-point Likert-scale items (1 = very low to 5 = very high). Interprofessional collaboration competencies were assessed using ICCAS, which is a validated instrument that measures self-reported competence across six domains including communication, collaboration, roles and responsibilities, collaborative patient- and family-centered approach, conflict management, and team functioning [27, 28]. Items are rated on a 5-point Likert scale, yielding a total score ranging from 20 to 100, with higher scores indicating greater perceived interprofessional competence. Although the ICCAS was originally developed as a retrospective tool asking students to rate their competence before and after the activity in a single post-test, it has also been validated for separate pre-test and post-test administration [27]. The ICCAS scale displayed high internal consistency with a Cronbach’s alpha coefficient of 0.79 at pre-test and 0.81 at post-test. Readiness and attitudes toward AI use in healthcare were evaluated using MAIRS that assesses four domains (cognition, ability, vision, and ethics) which reflect preparedness to engage with AI tools in clinical and educational contexts [29]. The MAIRS items are scored on a 5-point Likert scale, with higher scores indicating greater AI readiness. The MAIRS scale displayed high internal consistency with a Cronbach’s alpha coefficient of 0.85 at pre-test and 0.88 at post-test. AMS knowledge was assessed using a 10-item questionnaire developed by the researchers and focused on principles relevant to UTI management [25, 30]. The instrument consisted of two conceptual domains: core AMS principles (5 items) assessing general stewardship concepts such as appropriate antibiotic initiation, spectrum selection, reassessment based on microbiological results, and treatment duration; and UTI-specific stewardship application (5 items) evaluating clinical decision-making, including interpretation of urine cultures, differentiation between colonization and infection, empiric therapy selection, and de-escalation strategies. All items were formatted as true/false statements. Each correct response was awarded one point, yielding a total AMS score ranging from 0 to 10, with higher scores indicating greater stewardship knowledge. The AMS scale displayed high internal consistency with a Cronbach’s alpha coefficient of 0.83 at pre-test and 0.85 at post-test. The post-intervention survey included four open-ended questions designed to capture students’ reflective experiences. These questions explored students’ perceptions of interprofessional skills addressed during the simulation, the influence of AI-powered clinical avatars on clinical reasoning compared with traditional classroom discussions, perceived limitations or risks associated with reliance on AI for clinical or interprofessional guidance, and insights related to AMS responsibilities and decision-making during the simulation. Statistical analysis Data analysis was performed using IBM SPSS Statistics version 26.0. Descriptive statistics were presented as mean ± standard deviation (SD) for continuous variables and frequencies with percentages for categorical variables. Normality of distribution was assessed using the Shapiro-Wilk test and visual inspection of Q-Q plots. Cronbach’s alpha was used to test for the internal consistency of the scales used. Changes in outcome measures (ICCAS, MAIRS, AMS) from pre- to post-intervention were analyzed using paired-sample t-tests. Effect sizes for paired pre-post comparisons were calculated using Cohen’s d for repeated measures (dz). Effect sizes were interpreted as small (d = 0.2), medium (d = 0.5), or large (d = 0.8). To examine the overall effect of the intervention simultaneously, repeated-measures multivariate analysis of covariance (MANCOVA) was conducted, with time (pre vs. post) as the within-subject factor. Age, GPA, perceived digital literacy, familiarity with AI in healthcare, perceived usefulness of AI, and prior AI training were included as covariates. The assumption of homogeneity of regression slopes was tested by examining the interaction between time and each covariate. Interaction terms were included to explore differential effects by academic year, prior IPE exposure, and prior simulation experience. Wilks’ Lambda was used to assess multivariate significance, and effect sizes were expressed using partial eta-squared (η²). Significant multivariate findings were followed by univariate analyses and post-hoc comparisons with Bonferroni correction. An exploratory subgroup analysis compared students with and without prior IPE exposure. Baseline differences were assessed using independent t-tests, while within-group pre-post changes were evaluated using paired t-tests. Change scores (Δ = post - pre) were compared between groups using independent t-tests, with Δ differences reported alongside 95% confidence intervals and Cohen’s d effect sizes, calculated using the pooled standard deviation method. Multiple linear regression analyses were conducted to identify factors associated with post-intervention scores while controlling for baseline values. Variables with p < 0.2 in bivariate analyses were included, and results were reported as regression coefficients (B) with 95% confidence intervals and p-values, with significance set at p < 0.05. Given the pilot nature of the study, the census-based sampling approach, and the absence of a control group, all subgroup and multivariable analyses should be interpreted as exploratory and hypothesis-generating. Qualitative analysis Qualitative data from open-ended student reflections were analyzed using an inductive thematic analysis approach. Two members of the research team independently reviewed all responses to achieve data familiarization and conducted open inductive coding to identify meaningful units of text related to interprofessional learning, clinical reasoning, AMS, and perceptions of AI use. Coding was performed manually without qualitative analysis software due to the focused scope and manageable volume of responses. To enhance analytic rigor, two coders independently coded a random 20% sample (n = 22), with discrepancies resolved through discussion and consensus. Related codes were then grouped into higher-order themes, which were reviewed and refined in relation to the study objectives and underlying competency frameworks. Representative quotations were selected using a maximum variation approach to reflect diverse perspectives across academic years and thematic categories while maintaining participant anonymity. Results Demographic and educational characteristics of the students A total of 112 students participated in the study, with a mean age of 22.61 ± 2.15 years. Females constituted 73.2% of the sample. The mean GPA was 3.27 ± 0.39, and participants were distributed across third (35.7%), fourth (37.5%), and fifth (26.8%) academic years. Most students reported no prior formal clinical training in AMS (62.5%), no previous participation in IPE activities (64.3%), and no prior exposure to clinical simulation (58.0%). Formal training in AI was limited, with 79.5% reporting no prior AI course or workshop. Nevertheless, 79.5% of participants reported at least occasional prior use of AI tools, most commonly large language models such as ChatGPT (88.4%) followed by AI chatbots for education (40.2%). Baseline perceptions indicated moderate familiarity with AI in healthcare (mean 3.21 ± 0.76), moderate perceived usefulness of AI for learning (3.50 ± 0.79), and slightly lower perceived digital literacy (3.06 ± 0.88) on a 5-point scale. Table 1 demonstrates the characteristics of the studied participants. Table 1Sociodemographic and educational characteristics of the participantsCharacteristicTotal n = 112Age, years, mean ± SD22.61 ± 2.15Gender, n (%) Male30 (26.8) Female82 (73.2)GPA, mean ± SD3.27 ± 0.39Academic year, n (%) Third year40 (35.7) Fourth year42 (37.5) Fifth year30 (26.8)Prior clinical training on antimicrobial stewardship, n (%) No70 (62.5) Yes42 (37.5)Previous participation in IPE activities, n (%) No72 (64.3) Yes40 (35.7)Previous participation in clinical simulation activities, n (%) No65 (58.0) Yes47 (42.0)Prior attendance of AI course or workshop, n (%) No89 (79.5) Yes23 (20.5)AI tools used previously in learning or clinical reasoning, n (%) Never23 (20.5) Sometimes68 (60.7) Regularly21 (18.8)AI tools exposure and utilization, n (%) ChatGPT or other LLM99 (88.4) Clinical decision support system29 (25.9) Virtual patient simulation37 (33.0) Pharmacy dispensing automation27 (24.1) Drug-drug interaction or medication review AI40 (35.7) Clinical symptom checker29 (25.9) AI chatbots for education45 (40.2) AI conversation partners (healthcare)24 (21.4) Others25 (22.3)Familiarity with AI in Healthcare, mean ± SDRange (1–5)3.21 ± 0.76Perceived usefulness of AI for learning, mean ± SDRange (1–5)3.50 ± 0.79Perceived digital literacy, mean ± SDRange (1–5)3.06 ± 0.88 Table 1Sociodemographic and educational characteristics of the participantsCharacteristicTotal n = 112Age, years, mean ± SD22.61 ± 2.15Gender, n (%) Male30 (26.8) Female82 (73.2)GPA, mean ± SD3.27 ± 0.39Academic year, n (%) Third year40 (35.7) Fourth year42 (37.5) Fifth year30 (26.8)Prior clinical training on antimicrobial stewardship, n (%) No70 (62.5) Yes42 (37.5)Previous participation in IPE activities, n (%) No72 (64.3) Yes40 (35.7)Previous participation in clinical simulation activities, n (%) No65 (58.0) Yes47 (42.0)Prior attendance of AI course or workshop, n (%) No89 (79.5) Yes23 (20.5)AI tools used previously in learning or clinical reasoning, n (%) Never23 (20.5) Sometimes68 (60.7) Regularly21 (18.8)AI tools exposure and utilization, n (%) ChatGPT or other LLM99 (88.4) Clinical decision support system29 (25.9) Virtual patient simulation37 (33.0) Pharmacy dispensing automation27 (24.1) Drug-drug interaction or medication review AI40 (35.7) Clinical symptom checker29 (25.9) AI chatbots for education45 (40.2) AI conversation partners (healthcare)24 (21.4) Others25 (22.3)Familiarity with AI in Healthcare, mean ± SDRange (1–5)3.21 ± 0.76Perceived usefulness of AI for learning, mean ± SDRange (1–5)3.50 ± 0.79Perceived digital literacy, mean ± SDRange (1–5)3.06 ± 0.88 Sociodemographic and educational characteristics of the participants Familiarity with AI in Healthcare, mean ± SD Range (1–5) Perceived usefulness of AI for learning, mean ± SD Range (1–5) Perceived digital literacy, mean ± SD Range (1–5) Intervention effects on interprofessional collaboration, AI readiness, and antimicrobial stewardship outcomes Participants reported significantly higher post-intervention scores across all outcomes. The total ICCAS score increased significantly from 58.82 ± 14.67 to 68.52 ± 12.45 (p < 0.001, Cohen’s dz = 0.79), and across its subscales including communication, collaboration, roles and responsibilities, family and patient-centered care, conflict management, and team functioning (p < 0.001). Similarly, overall MAIRS scores increased significantly from 75.16 ± 10.85 to 82.12 ± 9.49 (p < 0.001, Cohen’s dz = 0.76), and across cognition, ability, vision, and ethics subdomains (p < 0.001). AMS knowledge had a large gain, with total AMS scores increasing from 6.21 ± 1.74 to 8.08 ± 1.35 (p < 0.001, Cohen’s dz = 1.34), driven by significant increase in both core stewardship principles and UTI-specific clinical application subscales (p < 0.001). Table 2 compares the ICCAS, MAIRS, and AMS scales and subscales before and after the activity. Table 2Comparison of the scales and subscales before and after the AI-enhanced simulated interprofessional learning interventionRangeMin-MaxMean ± SDMean change (post-pre)Cohen’s dzp-valuePre-activityPost-activityICCAS Scale Total ICCAS score20–10058.82 ± 14.6768.52 ± 12.45+ 9.700.79< 0.001 Communication subscale5–2514.38 ± 3.9016.74 ± 3.59+ 2.350.70< 0.001 Collaboration subscale3–158.67 ± 2.2410.15 ± 2.33+ 1.470.73< 0.001 Roles and responsibilities subscale4–2011.48 ± 3.1312.98 ± 3.06+ 1.500.54< 0.001 Collaborative patient/family-centered approach subscale3–159.05 ± 2.3510.53 ± 2.28+ 1.480.72< 0.001 Conflict management/resolution subscale3–158.99 ± 2.6010.70 ± 2.29+ 1.710.78< 0.001 Team functioning subscale2–106.24 ± 1.717.41 ± 1.64+ 1.170.78< 0.001MAIRS Scale Total MAIRS score22–11075.16 ± 10.8582.12 ± 9.49+ 6.960.76< 0.001 Cognition subscale8–4026.78 ± 3.9428.91 ± 3.91+ 2.120.60< 0.001 Ability subscale8–4027.87 ± 4.3430.24 ± 3.86+ 2.360.65< 0.001 Vision subscale3–1510.05 ± 2.1411.31 ± 2.03+ 1.260.67< 0.001 Ethics subscale3–1510.45 ± 2.3811.66 ± 2.15+ 1.210.59< 0.001AMS Scale Total AMS score0–106.21 ± 1.748.08 ± 1.35+ 1.861.34< 0.001 Core AMS principles subscale0–53.45 ± 1.154.26 ± 0.86+ 0.810.90< 0.001 UTI-specific AMS clinical application subscale0–52.76 ± 1.263.82 ± 0.92+ 1.051.07< 0.001Cohen’s dz was calculated using the standard deviation of the change scores for paired observations Table 2Comparison of the scales and subscales before and after the AI-enhanced simulated interprofessional learning interventionRangeMin-MaxMean ± SDMean change (post-pre)Cohen’s dzp-valuePre-activityPost-activityICCAS Scale Total ICCAS score20–10058.82 ± 14.6768.52 ± 12.45+ 9.700.79< 0.001 Communication subscale5–2514.38 ± 3.9016.74 ± 3.59+ 2.350.70< 0.001 Collaboration subscale3–158.67 ± 2.2410.15 ± 2.33+ 1.470.73< 0.001 Roles and responsibilities subscale4–2011.48 ± 3.1312.98 ± 3.06+ 1.500.54< 0.001 Collaborative patient/family-centered approach subscale3–159.05 ± 2.3510.53 ± 2.28+ 1.480.72< 0.001 Conflict management/resolution subscale3–158.99 ± 2.6010.70 ± 2.29+ 1.710.78< 0.001 Team functioning subscale2–106.24 ± 1.717.41 ± 1.64+ 1.170.78< 0.001MAIRS Scale Total MAIRS score22–11075.16 ± 10.8582.12 ± 9.49+ 6.960.76< 0.001 Cognition subscale8–4026.78 ± 3.9428.91 ± 3.91+ 2.120.60< 0.001 Ability subscale8–4027.87 ± 4.3430.24 ± 3.86+ 2.360.65< 0.001 Vision subscale3–1510.05 ± 2.1411.31 ± 2.03+ 1.260.67< 0.001 Ethics subscale3–1510.45 ± 2.3811.66 ± 2.15+ 1.210.59< 0.001AMS Scale Total AMS score0–106.21 ± 1.748.08 ± 1.35+ 1.861.34< 0.001 Core AMS principles subscale0–53.45 ± 1.154.26 ± 0.86+ 0.810.90< 0.001 UTI-specific AMS clinical application subscale0–52.76 ± 1.263.82 ± 0.92+ 1.051.07< 0.001Cohen’s dz was calculated using the standard deviation of the change scores for paired observations Comparison of the scales and subscales before and after the AI-enhanced simulated interprofessional learning intervention Cohen’s dz was calculated using the standard deviation of the change scores for paired observations Repeated-measures MANCOVA (Table 3) revealed a significant overall effect of time on the combined outcome measures following the participation in the AI-enhanced simulated interprofessional learning intervention (Wilks’ Λ = 0.412, F (3,98) = 46.52, p < 0.001), with a large multivariate effect size (partial η² = 0.588), after adjustment for age, GPA, digital literacy, AI familiarity, perceived usefulness of AI, and prior AI training. A significant interaction between time and academic year was observed (p = 0.013), suggesting variation in pre-post score changes across student levels, whereas interactions with prior IPE participation and prior simulation experience were not statistically significant. Fifth-year students had larger pre-post score differences across the three outcomes compared with third-year students, with effect sizes ranging from moderate to large (d = 0.62 to 0.78). Fourth-year students also had higher pre-post score differences relative to third-year students (d = 0.32 to 0.38). Comparisons between fifth- and fourth-year students showed smaller but statistically significant differences in ICCAS (p = 0.006, d = 0.48) and MAIRS (p = 0.029, d = 0.3) change scores, whereas the difference for AMS scores was not statistically significant (p = 0.162, d = 0.24), as presented in Table 4. Table 3Repeated-measures MANCOVA examining the overall effect of the AI-enhanced simulated interprofessional learning interventionEffectValueF (df1, df2)p-valuePartial η²Time (pre vs. post)0.41246.52 (3, 98)< 0.0010.588Time × Academic year0.8512.78 (6, 194)0.0130.079Time × Prior IPE participation0.9621.28 (3, 98)0.2870.038Time × Prior simulation experience0.9412.02 (3, 98)0.1160.059 Table 3Repeated-measures MANCOVA examining the overall effect of the AI-enhanced simulated interprofessional learning interventionEffectValueF (df1, df2)p-valuePartial η²Time (pre vs. post)0.41246.52 (3, 98)< 0.0010.588Time × Academic year0.8512.78 (6, 194)0.0130.079Time × Prior IPE participation0.9621.28 (3, 98)0.2870.038Time × Prior simulation experience0.9412.02 (3, 98)0.1160.059 Repeated-measures MANCOVA examining the overall effect of the AI-enhanced simulated interprofessional learning intervention Table 4Pairwise comparisons of mean change scores (Post-Pre) Between Academic YearsComparisonChange score (post-pre)95% CIp-value*Cohen’s dICCAS total Fourth vs. third year2.67[0.48, 4.86]0.0180.38 Fifth vs. third year6.60[3.85, 9.25]< 0.0010.78 Fifth vs. fourth year3.39[1.15, 5.71]0.0060.48MAIRS total Fourth vs. third year2.22[0.18, 4.26]0.0340.32 Fifth vs. third year4.95[2.42, 7.48]< 0.0010.65 Fifth vs. fourth year2.73[0.28, 5.18]0.0290.3AMS total Fourth vs. third year0.56[0.06, 1.06]0.0280.36 Fifth vs. third year0.93[0.42, 1.44]< 0.0010.62 Fifth vs. fourth year0.37[-0.15, 0.89]0.1620.24*p-values adjusted using Bonferroni correction for three pairwise comparisons per outcome. Cohen’s d calculated using pooled standard deviation of change scores for independent groups Table 4Pairwise comparisons of mean change scores (Post-Pre) Between Academic YearsComparisonChange score (post-pre)95% CIp-value*Cohen’s dICCAS total Fourth vs. third year2.67[0.48, 4.86]0.0180.38 Fifth vs. third year6.60[3.85, 9.25]< 0.0010.78 Fifth vs. fourth year3.39[1.15, 5.71]0.0060.48MAIRS total Fourth vs. third year2.22[0.18, 4.26]0.0340.32 Fifth vs. third year4.95[2.42, 7.48]< 0.0010.65 Fifth vs. fourth year2.73[0.28, 5.18]0.0290.3AMS total Fourth vs. third year0.56[0.06, 1.06]0.0280.36 Fifth vs. third year0.93[0.42, 1.44]< 0.0010.62 Fifth vs. fourth year0.37[-0.15, 0.89]0.1620.24*p-values adjusted using Bonferroni correction for three pairwise comparisons per outcome. Cohen’s d calculated using pooled standard deviation of change scores for independent groups Pairwise comparisons of mean change scores (Post-Pre) Between Academic Years *p-values adjusted using Bonferroni correction for three pairwise comparisons per outcome. Cohen’s d calculated using pooled standard deviation of change scores for independent groups Follow-up univariate analyses (Supplementary Table 2) revealed statistically significant pre-post differences across all outcomes, with adjusted post-intervention means remaining higher than adjusted pre-intervention values after controlling for covariates. Factors associated with post-intervention interprofessional, AI readiness, and antimicrobial stewardship outcomes Bivariate analyses were first conducted to explore associations between independent variables and post-intervention outcomes (Supplementary Table 3). Multivariable linear regression analyses (Table 5) explored factors associated with post-intervention outcomes after adjustment for baseline (pre-intervention) scores and relevant covariates. Compared with third-year students, fourth-year students had significantly higher post-intervention ICCAS (B = 3.45, 95% CI 0.65–6.25), MAIRS (B = 2.65, 95% CI 0.25–5.05), and AMS scores (B = 0.62, 95% CI 0.08–1.16). Similarly, fifth-year students had higher post-intervention ICCAS (B = 7.15, 95% CI 3.95–10.35), MAIRS (B = 5.15, 95% CI 2.45–7.85), and AMS (B = 1.08, 95% CI 0.48–1.68). Additionally, prior participation in IPE activities was significantly associated with higher post-intervention ICCAS (B = 2.95, 95% CI 0.15–5.75), MAIRS (B = 2.25, 95% CI 0.15–4.35), and AMS scores (B = 0.52, 95% CI 0.05–0.99). Also, prior clinical simulation experience was associated with higher ICCAS (B = 2.65, 95% CI 0.15–5.15) and MAIRS scores (B = 2.05, 95% CI 0.15–3.95), but not AMS knowledge. Moreover, regular prior use of AI tools was associated with higher ICCAS (B = 2.82, 95% CI 0.15–5.49) and MAIRS scores (B = 2.32, 95% CI 0.15–4.49). Higher perceived digital literacy was associated with higher ICCAS (B = 1.75, 95% CI 0.45–3.05), MAIRS (B = 1.45, 95% CI 0.38–2.52), and AMS (B = 0.30, 95% CI 0.05–0.55), whereas AI familiarity and perceived usefulness of AI were not significantly associated with outcomes after adjustment. Table 5Multivariable linear regression identifying independent factors of post-intervention ICCAS, MAIRS, and AMS scoresCharacteristicICCAS post-activity scoreMAIRS post-activity scoreAMS post-activity scoreB [95% CI]p-valueB [95% CI]p-valueB [95% CI]p-valueAcademic year* Fourth vs. third year3.45 [0.65, 6.25]0.0172.65 [0.25, 5.05]0.0310.62 [0.08, 1.16]0.025 Fifth vs. third year7.15 [3.95, 10.35]< 0.0015.15 [2.45, 7.85]< 0.0011.08 [0.48, 1.68]< 0.001Previous participation in IPE activities* (yes vs. no)2.95 [0.15, 5.75]0.0392.25 [0.15, 4.35]0.0360.52 [0.05, 0.99]0.031Previous participation in clinical simulation activities* (yes vs. no)2.65 [0.15, 5.15]0.0382.05 [0.15, 3.95]0.0350.45 [-0.02, 0.92]0.062AI tools used previously in learning or clinical reasoning* Sometimes vs. Never1.72 [-1.05, 4.49]0.2221.42 [-0.95, 3.79]0.1380.26 [-0.25, 0.77]0.315 Regularly vs. Never2.82 [0.15, 5.49]0.0392.32 [0.15, 4.49]0.0270.42 [-0.09, 0.93]0.105Familiarity with AI in Healthcare0.82 [-0.48, 2.12]0.2150.68 [-0.39, 1.75]0.2120.14 [-0.11, 0.39]0.275Perceived usefulness of AI for learning0.95 [-0.35, 2.25]0.1520.82 [-0.25, 1.89]0.1320.17 [-0.08, 0.42]0.185Perceived digital literacy1.75 [0.45, 3.05]0.0091.45 [0.38, 2.52]0.0080.30 [0.05, 0.55]0.019ICCAS pre-activity score0.42 [0.28, 0.56]< 0.0010.15 [-0.01, 0.31]0.0680.10 [-0.02, 0.22]0.105MAIRS pre-activity score0.12 [-0.04, 0.28]0.1320.45 [0.29, 0.61]< 0.0010.08 [-0.04, 0.20]0.185AMS pre-activity score0.18 [0.04, 0.32]0.0120.10 [-0.06, 0.26]0.2250.35 [0.23, 0.47]< 0.001*Reference categories: third year (for academic year), no prior IPE, no prior simulation, never used AI tools Table 5Multivariable linear regression identifying independent factors of post-intervention ICCAS, MAIRS, and AMS scoresCharacteristicICCAS post-activity scoreMAIRS post-activity scoreAMS post-activity scoreB [95% CI]p-valueB [95% CI]p-valueB [95% CI]p-valueAcademic year* Fourth vs. third year3.45 [0.65, 6.25]0.0172.65 [0.25, 5.05]0.0310.62 [0.08, 1.16]0.025 Fifth vs. third year7.15 [3.95, 10.35]< 0.0015.15 [2.45, 7.85]< 0.0011.08 [0.48, 1.68]< 0.001Previous participation in IPE activities* (yes vs. no)2.95 [0.15, 5.75]0.0392.25 [0.15, 4.35]0.0360.52 [0.05, 0.99]0.031Previous participation in clinical simulation activities* (yes vs. no)2.65 [0.15, 5.15]0.0382.05 [0.15, 3.95]0.0350.45 [-0.02, 0.92]0.062AI tools used previously in learning or clinical reasoning* Sometimes vs. Never1.72 [-1.05, 4.49]0.2221.42 [-0.95, 3.79]0.1380.26 [-0.25, 0.77]0.315 Regularly vs. Never2.82 [0.15, 5.49]0.0392.32 [0.15, 4.49]0.0270.42 [-0.09, 0.93]0.105Familiarity with AI in Healthcare0.82 [-0.48, 2.12]0.2150.68 [-0.39, 1.75]0.2120.14 [-0.11, 0.39]0.275Perceived usefulness of AI for learning0.95 [-0.35, 2.25]0.1520.82 [-0.25, 1.89]0.1320.17 [-0.08, 0.42]0.185Perceived digital literacy1.75 [0.45, 3.05]0.0091.45 [0.38, 2.52]0.0080.30 [0.05, 0.55]0.019ICCAS pre-activity score0.42 [0.28, 0.56]< 0.0010.15 [-0.01, 0.31]0.0680.10 [-0.02, 0.22]0.105MAIRS pre-activity score0.12 [-0.04, 0.28]0.1320.45 [0.29, 0.61]< 0.0010.08 [-0.04, 0.20]0.185AMS pre-activity score0.18 [0.04, 0.32]0.0120.10 [-0.06, 0.26]0.2250.35 [0.23, 0.47]< 0.001*Reference categories: third year (for academic year), no prior IPE, no prior simulation, never used AI tools Multivariable linear regression identifying independent factors of post-intervention ICCAS, MAIRS, and AMS scores *Reference categories: third year (for academic year), no prior IPE, no prior simulation, never used AI tools Exploratory subgroup analysis of pre-post score differences according to prior IPE exposure At baseline, students with prior IPE exposure had significantly higher ICCAS scores (mean difference = 10.10, 95% CI: 4.85 to 15.35, p = 0.002, Cohen’s d = 0.75) and MAIRS scores (mean difference = 6.70, 95% CI: 2.45 to 10.95, p = 0.008, Cohen’s d = 0.64) compared to students without prior IPE exposure. However, baseline AMS knowledge scores were similar between the two groups (mean difference = 0.30, 95% CI: -0.35 to 0.95, p = 0.412, Cohen’s d = 0.18). Both groups had statistically significant pre-post score differences across all outcomes (all p < 0.001). Students without prior IPE exposure showed larger ICCAS change scores (Δ = 11.20 ± 8.50) compared with students with prior IPE exposure (Δ = 7.00 ± 7.20). The between-group difference in ICCAS change scores was statistically significant (mean Δ difference = 4.20, 95% CI: 1.20 to 7.20, p = 0.008), with a moderate effect size (Cohen’s d = 0.52). Differences in MAIRS and AMS change scores between groups were not statistically significant (p > 0.05). Given the exploratory subgroup nature of these analyses and the absence of a control group, these findings should be interpreted cautiously. Table 6 summarizes the comparison of pre-post changes between students with and without prior IPE exposure. Table 6Comparison of pre-post changes between students with and without prior IPE exposureOutcomeNo Prior IPE (n = 72)Prior IPE (n = 40)Between-group comparisonPrePostΔ (change)p-valuePrePostΔ (change)p-valueΔ difference [95% CI]p-valueCohen's dCohen's dICCAS total score55.20 ± 14.5066.40 ± 12.8011.20 ± 8.50< 0.00165.30 ± 12.1072.30 ± 11.207.00 ± 7.20< 0.0014.20 [1.20, 7.20]0.0080.52MAIRS total score72.80 ± 10.9080.90 ± 9.608.10 ± 7.80< 0.00179.50 ± 9.8086.20 ± 8.906.70 ± 6.90< 0.0011.4 [-1.42, 4.20]0.3290.19AMS total score6.10 ± 1.808.05 ± 1.401.95 ± 1.30< 0.0016.40 ± 1.608.15 ± 1.251.75 ± 1.20< 0.0010.20 [-0.28, 0.67]0.4140.16Δ = change score (post − pre). Δ difference = (Δ_no IPE) − (Δ_prior IPE). Cohen’s d calculated using pooled standard deviation of change scores for independent groups Table 6Comparison of pre-post changes between students with and without prior IPE exposureOutcomeNo Prior IPE (n = 72)Prior IPE (n = 40)Between-group comparisonPrePostΔ (change)p-valuePrePostΔ (change)p-valueΔ difference [95% CI]p-valueCohen's dCohen's dICCAS total score55.20 ± 14.5066.40 ± 12.8011.20 ± 8.50< 0.00165.30 ± 12.1072.30 ± 11.207.00 ± 7.20< 0.0014.20 [1.20, 7.20]0.0080.52MAIRS total score72.80 ± 10.9080.90 ± 9.608.10 ± 7.80< 0.00179.50 ± 9.8086.20 ± 8.906.70 ± 6.90< 0.0011.4 [-1.42, 4.20]0.3290.19AMS total score6.10 ± 1.808.05 ± 1.401.95 ± 1.30< 0.0016.40 ± 1.608.15 ± 1.251.75 ± 1.20< 0.0010.20 [-0.28, 0.67]0.4140.16Δ = change score (post − pre). Δ difference = (Δ_no IPE) − (Δ_prior IPE). Cohen’s d calculated using pooled standard deviation of change scores for independent groups Comparison of pre-post changes between students with and without prior IPE exposure Δ = change score (post − pre). Δ difference = (Δ_no IPE) − (Δ_prior IPE). Cohen’s d calculated using pooled standard deviation of change scores for independent groups Integrated quantitative and qualitative findings from student reflections on the AI-enhanced simulated interprofessional learning intervention Integration of quantitative and qualitative findings showed convergence across several outcome domains. Students’ reflections complemented the observed higher post-intervention ICCAS scores by describing perceived improvements in interprofessional communication, teamwork, role clarification, and collaborative discussion during the simulated activity. Participants frequently reported greater confidence in articulating clinical reasoning, actively listening to team perspectives, and collaborating under simulated time constraints. Similarly, the higher post-intervention MAIRS scores were complemented by students’ descriptions of active engagement with AI-supported clinical reasoning, stepwise therapeutic decision-making, and critical appraisal of AI-generated responses. Students also reported awareness of ethical and practical limitations associated with AI use in healthcare, including concerns regarding over-reliance, incomplete contextual understanding, lack of empathy, and the continued importance of clinician oversight and guideline verification. In addition, the observed higher AMS scores aligned with students’ reflections regarding antimicrobial stewardship responsibilities, including greater attention to appropriate antimicrobial selection, avoidance of unnecessary antibiotic use, accountability in prescribing decisions, and guideline-informed therapeutic reasoning. Integrated qualitative themes, representative quotations, and their relationship to quantitative findings are summarized in Table 7. Table 7Integration of quantitative and qualitative findingsQuantitative OutcomeIntegrated Qualitative ThemeIntegrated Qualitative SubthemesRepresentative Student QuotesIntegrated InterpretationHigher post-intervention ICCAS scoresPerceived development of interprofessional communication, teamwork, and role clarification• Communication• Teamwork• Role clarification• Working under pressure“The simulation required me to clearly share my reasoning with the team and listen to others’ perspectives, which strengthened teamwork and communication.” (Fifth-year student)“I improved my communication skills by practicing clear, structured updates and actively listening, which made teamwork smoother during the simulation.” (Fourth-year student)Students’ reflections complemented higher self-reported ICCAS scores by describing perceived improvements in communication, teamwork, collaborative discussion, and understanding of professional roles during the simulated interprofessional activity.Higher post-intervention MAIRS scoresActive AI-supported clinical reasoning and critical awareness of AI limitations• Step-by-step decision-making• Application of guidelines• Pattern recognition• Realistic clinical pressure• Risk of over-reliance• Oversimplification• Lack of empathy• Need for human oversight and guideline verification“Unlike classroom discussion, the avatars forced me to think step-by-step, justify each decision, and adjust my plan in real time.” (Third-year student)“The AI avatars made my clinical reasoning more practical and patient-focused by requiring me to connect symptoms, history, and treatment options under pressure.” (Fourth-year student)“AI should be used as a supportive tool, not a decision-maker and clinical judgment and guidelines must always guide final decisions.” (Fifth-year student)“A key limitation is the risk of over-relying on AI without verifying accuracy, so in real practice I would always cross-check with guidelines and consult colleagues.” (Fourth-year student)Qualitative findings complemented higher MAIRS scores by demonstrating students’ perceived engagement with AI-supported reasoning while simultaneously recognizing limitations related to over-reliance, contextual judgment, and ethical use of AI technologies in healthcare.Higher post-intervention AMS scoresIncreased awareness of antimicrobial stewardship responsibilities and accountability• Appropriate antimicrobial selection• Avoidance of unnecessary antibiotic use• Guideline-based decision-making• Accountability for antimicrobial choices“The simulation helped me realize that choosing an antibiotic is not just about what works, but about whether it is truly necessary and appropriate for the patient.” (Third-year student)“Working through the case made me more careful about antibiotic use and reinforced the importance of following guidelines and discussing decisions with the team.” (Fifth-year student)Students’ reflections aligned with higher AMS scores by describing greater attention to stewardship principles, guideline-informed antimicrobial use, accountability for therapeutic decisions, and avoidance of unnecessary antibiotic prescribing. Table 7Integration of quantitative and qualitative findingsQuantitative OutcomeIntegrated Qualitative ThemeIntegrated Qualitative SubthemesRepresentative Student QuotesIntegrated InterpretationHigher post-intervention ICCAS scoresPerceived development of interprofessional communication, teamwork, and role clarification• Communication• Teamwork• Role clarification• Working under pressure“The simulation required me to clearly share my reasoning with the team and listen to others’ perspectives, which strengthened teamwork and communication.” (Fifth-year student)“I improved my communication skills by practicing clear, structured updates and actively listening, which made teamwork smoother during the simulation.” (Fourth-year student)Students’ reflections complemented higher self-reported ICCAS scores by describing perceived improvements in communication, teamwork, collaborative discussion, and understanding of professional roles during the simulated interprofessional activity.Higher post-intervention MAIRS scoresActive AI-supported clinical reasoning and critical awareness of AI limitations• Step-by-step decision-making• Application of guidelines• Pattern recognition• Realistic clinical pressure• Risk of over-reliance• Oversimplification• Lack of empathy• Need for human oversight and guideline verification“Unlike classroom discussion, the avatars forced me to think step-by-step, justify each decision, and adjust my plan in real time.” (Third-year student)“The AI avatars made my clinical reasoning more practical and patient-focused by requiring me to connect symptoms, history, and treatment options under pressure.” (Fourth-year student)“AI should be used as a supportive tool, not a decision-maker and clinical judgment and guidelines must always guide final decisions.” (Fifth-year student)“A key limitation is the risk of over-relying on AI without verifying accuracy, so in real practice I would always cross-check with guidelines and consult colleagues.” (Fourth-year student)Qualitative findings complemented higher MAIRS scores by demonstrating students’ perceived engagement with AI-supported reasoning while simultaneously recognizing limitations related to over-reliance, contextual judgment, and ethical use of AI technologies in healthcare.Higher post-intervention AMS scoresIncreased awareness of antimicrobial stewardship responsibilities and accountability• Appropriate antimicrobial selection• Avoidance of unnecessary antibiotic use• Guideline-based decision-making• Accountability for antimicrobial choices“The simulation helped me realize that choosing an antibiotic is not just about what works, but about whether it is truly necessary and appropriate for the patient.” (Third-year student)“Working through the case made me more careful about antibiotic use and reinforced the importance of following guidelines and discussing decisions with the team.” (Fifth-year student)Students’ reflections aligned with higher AMS scores by describing greater attention to stewardship principles, guideline-informed antimicrobial use, accountability for therapeutic decisions, and avoidance of unnecessary antibiotic prescribing. Integration of quantitative and qualitative findings • Communication • Teamwork • Role clarification • Working under pressure “The simulation required me to clearly share my reasoning with the team and listen to others’ perspectives, which strengthened teamwork and communication.” (Fifth-year student) “I improved my communication skills by practicing clear, structured updates and actively listening, which made teamwork smoother during the simulation.” (Fourth-year student) • Step-by-step decision-making • Application of guidelines • Pattern recognition • Realistic clinical pressure • Risk of over-reliance • Oversimplification • Lack of empathy • Need for human oversight and guideline verification “Unlike classroom discussion, the avatars forced me to think step-by-step, justify each decision, and adjust my plan in real time.” (Third-year student) “The AI avatars made my clinical reasoning more practical and patient-focused by requiring me to connect symptoms, history, and treatment options under pressure.” (Fourth-year student) “AI should be used as a supportive tool, not a decision-maker and clinical judgment and guidelines must always guide final decisions.” (Fifth-year student) “A key limitation is the risk of over-relying on AI without verifying accuracy, so in real practice I would always cross-check with guidelines and consult colleagues.” (Fourth-year student) • Appropriate antimicrobial selection • Avoidance of unnecessary antibiotic use • Guideline-based decision-making • Accountability for antimicrobial choices “The simulation helped me realize that choosing an antibiotic is not just about what works, but about whether it is truly necessary and appropriate for the patient.” (Third-year student) “Working through the case made me more careful about antibiotic use and reinforced the importance of following guidelines and discussing decisions with the team.” (Fifth-year student) Discussion This pilot study provides preliminary evidence suggesting that participation in an AI-enhanced simulated interprofessional learning activity was associated with higher post-intervention self-reported interprofessional collaboration competencies, perceived readiness for AI use in healthcare, and short-term AMS knowledge scores among pharmacy students. Qualitative findings complemented the quantitative results. The findings suggest that AI-enhanced simulated interprofessional learning may represent a feasible and potentially valuable educational approach for supporting perceived interprofessional competence, perceived AI readiness, and short-term AMS knowledge gains within competency-based pharmacy education. The higher post-intervention ICCAS scores observed in this pilot study are relevant to the expanding role of pharmacists as members of multidisciplinary healthcare teams. Pharmacists are increasingly involved in direct patient care, therapeutic decision-making, and medication safety, particularly in complex clinical contexts such as infectious diseases and AMS [31]. Previous studies have reported positive associations between simulation-based interprofessional learning and perceived teamwork, communication, and role clarity among healthcare students [32, 33]. For example, Sezgin and Bektas [12] reported higher teamwork and communication outcomes across randomized trials, while Saragih et al. [32] and Marion-Martins and Pinho [33] described improvements in interprofessional knowledge and collaborative competencies following simulation-based educational interventions. In pharmacy education, simulation-based IPE has similarly been associated with enhanced engagement in team-based care and communication of pharmacotherapeutic recommendations [34]. The present findings add preliminary exploratory evidence suggesting that AI-enhanced simulated interprofessional role interaction may facilitate exposure to key components of interprofessional learning, even in the absence of learners from multiple disciplines. Students’ qualitative reflections describing communication, teamwork, and role clarification experiences further complement the quantitative findings. In addition, participants reported perceived engagement in pharmacist-related interprofessional activities such as discussing medication-related decisions, contributing to treatment planning, and participating in collaborative clinical reasoning processes [35, 36]. These observations may be particularly relevant in educational settings where logistical barriers, including scheduling constraints and limited resources, challenge implementation of traditional interprofessional learning activities [37, 38]. Although statistically significant pre-post differences were observed across ICCAS subdomains, these findings should be interpreted cautiously given the exploratory pilot design, reliance on self-reported measures, and absence of a control group. Further controlled longitudinal studies using objective performance-based assessments are needed to better evaluate the educational impact of AI-enhanced simulated interprofessional learning on objectively demonstrated collaborative pharmacy practice competencies. The higher post-intervention MAIRS scores observed across cognition, ability, vision, and ethics domains may be relevant within the context of evolving pharmacy practice and the growing integration of AI technologies into healthcare. Although AI is increasingly recognized as a potentially transformative force in healthcare, its integration into health professions education remains limited and inconsistently implemented [39, 40]. Pharmacists are increasingly exposed to digital tools, clinical decision support systems, and AI-driven platforms to support medication management and patient safety [41]. In addition to technical use of these technologies, pharmacists are expected to critically appraise AI-generated outputs, assess their alignment with clinical guidelines, and integrate them appropriately within patient-centered decision-making processes. However, previous literature suggests that exposure to AI tools alone may not ensure appropriate competence and may contribute to over-reliance or uncritical acceptance of AI-generated information [42]. In the present study, higher post-intervention AI readiness scores were accompanied by qualitative reflections describing awareness of potential AI limitations, including perceived inaccuracies, limited contextual judgment, and insufficient patient-centered considerations. These qualitative observations align with emerging recommendations emphasizing that AI integration in healthcare education should incorporate critical appraisal, ethical reasoning, and human oversight [19, 42]. These findings provide preliminary support for the potential educational value of AI-enhanced simulation in promoting reflective engagement with AI technologies within pharmacy education. Importantly, the present study design does not permit isolation of the independent educational contribution of the AI component itself. The observed improvements may reflect the combined influence of multiple educational elements integrated within the intervention, including simulation-based learning, collaborative discussion, structured debriefing, case-based learning, and interprofessional role interaction. Because no comparator group involving non-AI simulation or traditional instruction was included, it remains unclear whether similar outcomes would have been achieved without AI-enhanced virtual avatars. Future controlled studies directly comparing AI-enhanced and non-AI simulation approaches are required to determine the incremental educational value attributable specifically to AI integration. The higher post-intervention AMS knowledge scores observed in this study may be relevant given the recognized role of pharmacists in antimicrobial stewardship programs. Importantly, the AMS assessment measured short-term knowledge score changes rather than demonstrated antimicrobial stewardship competence. Therefore, the observed AMS findings should be interpreted as short-term knowledge gains, not as definitive evidence of stewardship competency development or future prescribing performance. Pharmacists contribute to AMS through supporting appropriate antibiotic selection, dosing, duration, de-escalation, and education of healthcare professionals and patients [23]. Previous educational studies have reported positive associations between active learning approaches, including simulation and case-based learning, and AMS-related knowledge acquisition and prescribing awareness [43, 44]. In the present study, the observed pre-post differences in AMS scores may reflect exposure to multiple educational components incorporated within the intervention, including realistic clinical scenarios, interprofessional role interaction, and discussion of evidence-based therapeutic decision-making. Qualitative reflections further complemented these findings, with students describing increased attention to appropriate antimicrobial use, avoidance of unnecessary prescribing, and accountability in treatment-related decisions. These observations are generally consistent with competency domains commonly emphasized within antimicrobial stewardship education frameworks for pharmacists [25]. However, given the exploratory pilot design, absence of a control group, reliance on a researcher-developed AMS instrument, and short-term assessment period, these findings should be interpreted cautiously. Further controlled longitudinal studies using validated objective performance-based measures are needed to better evaluate the educational impact of AI-enhanced simulation on antimicrobial stewardship competencies. The qualitative findings provide additional context for interpreting the quantitative results. Students reported perceived experiences involving structured step-by-step clinical reasoning, integration of patient-specific information with guideline-based recommendations, and adaptation of decisions during simulated interactions. These processes are commonly associated with pharmacist participation in clinical decision-making and are consistent with prior simulation-based education literature describing immersive and interactive learning environments as supportive of learner engagement and perceived clinical reasoning development [34, 45]. Participants also perceived the AI-powered avatars as contributing to the realism and interactivity of the simulation by prompting ongoing discussion, justification, and reconsideration of pharmacotherapeutic decisions. Similar observations have been described in studies involving virtual patients and digital simulation platforms, which have reported perceived benefits related to interprofessional learning, role clarification, and collaborative decision-making experiences [46, 47]. These findings suggest that AI-enhanced simulation may represent a potentially flexible educational approach that warrants further evaluation in controlled educational settings. Importantly, students frequently acknowledged concerns regarding over-reliance on AI, limited contextual judgment, and the continued need for clinician oversight and guideline verification. These observations suggest that students did not perceive AI uncritically despite reporting higher self-reported readiness and engagement. The association between higher academic year and higher post-intervention scores may suggest that more advanced pharmacy students are better positioned to engage with complex simulation-based and interprofessional learning activities, potentially due to greater prior pharmacotherapeutic knowledge, clinical exposure, and professional socialization. This interpretation is generally consistent with educational literature suggesting that learners with stronger prior knowledge may be more able to engage in higher-order clinical reasoning and integration of multidisciplinary perspectives during simulation-based learning experiences [48, 49]. However, statistically significant pre-post score differences were observed across all academic levels, which may indicate that such educational activities can be adapted across different stages of pharmacy training. Higher perceived digital literacy was also associated with higher post-intervention scores across the measured outcomes, highlighting the potential relevance of digital competencies within contemporary pharmacy education. As pharmacists increasingly interact with digital technologies and AI-supported tools, the ability to critically appraise, interpret, and appropriately apply digital information may become increasingly important [47]. This observation is generally aligned with emerging frameworks describing digital health literacy as an important professional competency that extends beyond technical proficiency to include critical thinking, interpretation of digital information, and ethical use of technology [19, 50]. In contrast, AI familiarity and perceived usefulness of AI were not significantly associated with outcomes after adjustment, suggesting that prior exposure to AI tools alone may not necessarily translate into higher self-reported competencies or knowledge scores within educational contexts. These findings may support the importance of structured educational integration and guided reflective learning when incorporating AI technologies into pharmacy education [51]. Additionally, prior IPE and simulation exposure were associated with higher post-intervention scores across several outcomes. Although these findings should be interpreted cautiously given the exploratory nature of the analyses, they are generally consistent with previous literature suggesting that repeated exposure to collaborative and simulation-based educational experiences may support perceived development of teamwork, communication, and professional competencies over time [52, 53]. These observations may support further exploration of interprofessional, simulation-based, and digitally enhanced educational approaches across pharmacy curricula. The exploratory subgroup comparison between students with and without prior IPE exposure provided additional context for interpreting the observed pre-post score differences. Students without prior IPE exposure had lower self-reported baseline interprofessional competency scores but larger ICCAS change scores following the intervention compared with students who previously participated in IPE activities. One possible interpretation is that students with less prior exposure to collaborative learning experiences may have had greater opportunity for change during the activity. This observation is generally consistent with educational literature suggesting that learners with lower baseline scores may demonstrate larger relative score changes following structured educational experiences [48]. However, this finding should be interpreted cautiously, as the larger changes observed among students without prior IPE exposure may also partially reflect regression to the mean and differences in baseline scores between groups given the lower baseline scores within this subgroup. For perceived AI readiness and AMS knowledge, both groups showed statistically significant pre-post score differences, while between-group differences in change scores were not statistically significant. These findings may suggest that the observed score changes in these domains were not strongly dependent on prior IPE exposure. In addition, the absence of statistically significant differences between groups for MAIRS and AMS change scores may indicate that students with previous IPE exposure did not demonstrate clear ceiling effects within these measures. Importantly, given the absence of randomization, lack of a control group, exploratory subgroup design, and reliance on self-reported measures, these analyses do not permit causal inference regarding intervention effectiveness or subgroup responsiveness. Rather, these findings should be interpreted as preliminary and hypothesis-generating observations that may inform future controlled longitudinal studies evaluating AI-enhanced simulated interprofessional learning across learners with varying prior interprofessional educational exposure. Educational considerations and potential curricular implications The present pilot findings may have potential implications for pharmacy education, particularly in institutions operating across multiple campuses and serving large and diverse student populations, such as the Lebanese International University School of Pharmacy. Traditional interprofessional education models requiring synchronous coordination across multiple disciplines may present logistical and resource-related challenges. Within this context, AI-enhanced simulated interprofessional learning may represent a potentially flexible and standardized educational approach that could facilitate structured interprofessional interactions across distributed educational settings. Through AI-powered virtual avatars representing different healthcare roles, students may be exposed to simulated communication, teamwork, and collaborative clinical reasoning experiences without requiring simultaneous physical participation of multiple disciplines. Additional educational observations also emerged from this study. Students reported engaging with AI within structured simulation-based learning activities involving active participation, clinical reasoning, and reflective discussion. Rather than functioning solely as an information-delivery tool, the AI-supported simulation environment appeared to encourage students to discuss therapeutic decisions, reflect on AI-generated outputs, and consider evidence-based guideline alignment. These observations are generally consistent with emerging literature suggesting that AI integration in health professions education may be most valuable when incorporated within pedagogically guided learning environments that support higher-order thinking and critical appraisal rather than passive information retrieval [19]. Higher perceived digital literacy was also associated with higher post-intervention scores across several outcomes, suggesting that digital competencies may play an important role within contemporary pharmacy education. The findings may further suggest that familiarity with AI tools alone does not necessarily translate into higher perceived competencies or knowledge outcomes. Rather, structured educational integration and guided reflective engagement may be important considerations when incorporating AI technologies into pharmacy curricula. These exploratory findings may support further consideration of digital health literacy, AI awareness, and simulation-based interprofessional learning within pharmacy education frameworks. However, future controlled and longitudinal studies are needed before broader curricular recommendations can be made. Strengths and limitations This pilot study has several strengths relevant to pharmacy education and interprofessional learning research. The explanatory sequential mixed-methods design allowed integration of quantitative and qualitative findings, providing complementary insight into participants’ experiences and post-intervention outcomes. The use of previously validated instruments, including ICCAS and MAIRS, alongside assessment of multiple educational domains, enabled a multidimensional evaluation aligned with contemporary pharmacy education frameworks. In addition, the study explored a novel educational approach integrating AI-enhanced simulation, simulated interprofessional learning, and antimicrobial stewardship within a clinically relevant case-based context, addressing an area that remains relatively underexplored in the literature. Nevertheless, several limitations should be considered when interpreting the findings. The study employed a single-group pre-post quasi-experimental pilot design without a control group, limiting causal inference. The observed pre-post score differences may have been influenced by alternative explanations including testing effects, maturation, novelty effects related to AI technology exposure, Hawthorne effects, social desirability bias, or regression to the mean. The possibility of testing effects should be considered, particularly for the AMS knowledge assessment. The same questionnaire was administered at pre- and post-intervention, and the AMS instrument consisted of a relatively brief 10-item true/false scale closely aligned with the intervention content. Consequently, improved post-intervention AMS scores may partially reflect recall of prior responses, familiarity with questionnaire items, or independent review of relevant material rather than educational effects attributable solely to the intervention itself. Accordingly, interpretation of AMS score changes should remain cautious. Maturation effects should also be considered, as pharmacy students may naturally develop clinical reasoning abilities, interprofessional awareness, AI familiarity, and antimicrobial stewardship knowledge over the course of an academic semester independent of the intervention itself through ongoing coursework, clinical exposure, experiential training, and self-directed learning activities. Additionally, participants’ awareness of being observed within an educational research context may have contributed to altered engagement or response behavior (Hawthorne effect), potentially influencing post-intervention self-reported outcomes. Also, the larger ICCAS change scores observed among students without prior IPE exposure may partially reflect lower baseline scores and greater opportunity for score change upon repeated assessment, consistent with possible regression to the mean. Moreover, participation was voluntary and the response rate was relatively low. Consequently, self-selection bias cannot be excluded, as participating students may have been more motivated, more technologically engaged, or more interested in interprofessional and AI-enhanced learning than non-participants. Therefore, the observed outcomes may not fully reflect the experiences or perceptions of the broader pharmacy student population. Also, because the intervention incorporated multiple educational components, including simulation-based learning, AI-enhanced virtual avatars, collaborative teamwork, antimicrobial stewardship instruction, case-based learning, and facilitated debriefing, the study design does not permit isolation of the specific educational contribution attributable to the AI component itself. Future randomized controlled or comparative studies are needed to more rigorously evaluate the educational impact of AI-enhanced simulated interprofessional learning. Additionally, although exploratory subgroup analyses according to prior IPE exposure provided additional contextual insight, these analyses remain observational and non-randomized. Students with and without prior IPE exposure may differ in unmeasured characteristics such as motivation, academic engagement, or prior interest in collaborative learning. Furthermore, the sample size was based on the accessible student cohort participating in the educational activity rather than formal a priori power calculations. Although this pragmatic approach is common in educational intervention research, subgroup, interaction, and multivariable analyses should be interpreted as exploratory and hypothesis-generating rather than confirmatory. In addition, the study was conducted at a single institution and included only pharmacy students, which may limit transferability to other educational settings and healthcare disciplines. Although the intervention incorporated simulated interprofessional interaction through AI-powered avatars, participants did not engage with learners from other healthcare professions. Consequently, the interprofessional component reflected simulated rather than authentic interprofessional collaboration and may not fully capture the complexity and bidirectional dynamics of real-world team-based practice. Several outcomes, including interprofessional competencies and AI readiness, were assessed using self-reported instruments. Although validated scales were used, self-reported measures primarily reflect perceived competence and readiness rather than objectively demonstrated performance and may be subject to response bias, including social desirability and overestimation of competence. In addition, enthusiasm and novelty associated with exposure to AI-enhanced simulation may have contributed to higher post-intervention self-ratings independent of actual competency acquisition. Consequently, increases in perceived readiness or confidence should not be interpreted as definitive evidence of objectively demonstrated improvement in clinical or interprofessional performance. Future studies should incorporate objective performance-based assessments such as OSCEs, structured observational evaluations, or analysis of recorded team interactions using validated teamwork assessment frameworks. Moreover, outcome assessment was limited to the short-term post-intervention period, and therefore the durability and long-term educational impact of the observed findings remain uncertain. Longitudinal studies are needed to evaluate retention of competencies and potential translation into clinical or collaborative practice behaviors over time. Finally, although several factors were statistically associated with post-intervention outcomes, residual confounding cannot be excluded. Variables such as learner motivation, prior academic experiences, and learning preferences were not comprehensively assessed and may have influenced the findings. Furthermore, although subgroup comparisons, interaction analyses, and multivariable regression models were conducted to explore factors associated with the observed outcomes, the pilot sample size may have provided limited statistical power for these analyses. Consequently, these findings should be interpreted cautiously and viewed as exploratory and hypothesis-generating rather than confirmatory. Conclusion This pilot mixed-methods study provides preliminary evidence suggesting that participation in an AI-enhanced simulated interprofessional learning activity was associated with higher post-intervention self-reported interprofessional collaboration competencies, perceived AI readiness, and short-term AMS knowledge scores among pharmacy students. The integration of AI-supported simulation, interprofessional role interaction, and clinically contextualized learning activities may represent a potentially feasible educational approach within pharmacy education, particularly in settings where implementation of traditional interprofessional learning activities may be logistically challenging. Exploratory subgroup analyses suggested larger ICCAS change scores among students without prior IPE exposure; however, these findings should be interpreted cautiously given the absence of a control group and the possibility of regression to the mean. The observed associations between perceived digital literacy and post-intervention outcomes further highlight the potential relevance of digital competencies within contemporary pharmacy education. Nevertheless, the findings should be interpreted in light of the study’s exploratory pilot design, reliance on self-reported measures, single-institution setting, and short-term assessment period. Future randomized controlled and longitudinal studies incorporating multiple healthcare disciplines and objective performance-based assessments are needed to determine whether these perceived and short-term knowledge gains translate into demonstrated interprofessional, AI-related, or stewardship competencies. Supplementary Information Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2. Supplementary Material 3. Supplementary Material 3. Supplementary Material 4. Supplementary Material 4. Supplementary Material 5. Supplementary Material 5. Supplementary Material 6. Supplementary Material 6.