Artificial Intelligence Virtual Patient: A proof of concept study

Artificial Intelligence Virtual Patient: A proof of concept study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Artificial Intelligence Virtual Patient: A proof of concept study Betty S Chan, Timothy Dodds, Jen Xiang, Lucy Jo, Paul Dyer, Bhavani Kannan, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6272736/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 9 You are reading this latest preprint version Abstract Background Artificial Intelligence(AI) is advancing, but its role in simulating detailed patient-doctor interactions in the style of Objective Structured Clinical Examinations(OSCEs) is emerging. This study's goal was to create and validate an AI virtual patient(AIVP) that could interact with medical students, mimic a patient with a medical issue, and provide students feedback on their performance. Methods Six AIVP were developed to simulate OSCE scenarios for common emergency department presentations. The simulations were created using the Unity game engine, featuring a conversation loop that includes speech-to-text conversion (OpenAI Whisper), response generation(Open AI ChatGPT 4o), and speech generation (OpenAI TTS). A tutor AI(ChatGPT 4o) then generates feedback after the conversation to help students improve their responses. Final-year medical students were given the opportunity to interact with the AIVPs and participated in pre- and post-AIVP OSCE assessments to evaluate the AIVP's effect on performance, with Wilcoxon paired t-tests used for analysis. Students completed Likert Scales and surveys on the AIVP's educational value and technical issues. Results Twenty-one students participated over two weeks for a total of 21.7 hours, averaging 1.1 hour per user. The median OSCE scores improved from 63/100 (IQR: 53.5–70) to 70/100 (IQR: 63-73.5) (p = 0.29). On a Likert scale of 0 (strongly disagree) to 5 (strongly agree) there was strong agreement that the AIVP was a valuable learning experience(mean 4.62, SD 0.65). Students valued the feedback provided by the AIVP at the end of their interaction on their performance(mean 4.38, SD 0.84), Technical issues like voice recognition problems, latency in AIVP interaction, and occasional role reversals were reported. Conclusion This is a novel tool for developing history-taking skills and OSCE performance. Students found their interactions with the AIVP, and the feedback it provided on their performance, to be a valuable learning experience. However, technical factors and AIVP realism need further development. Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Artificial Intelligence (AI) has the potential to recreate high-fidelity human interactions with users. Early examples in academic literature demonstrate AI being trained to assume the role of a patient in a medical setting( 1 ). This is a rapidly evolving technology, and its application in medical education is still in its infancy. In medicine, taking a medical history is a fundamental skill for any medical professional. It is estimated that 80% of diagnoses can be made from history alone( 2 ). This process involves asking the patient a series of questions to gather sufficient, holistic information to make a provisional diagnosis. Diagnostic reasoning then follows, along with an explanation to the patient about the next steps in assessment, investigation, and treatment. However, finding patients in hospitals for history-taking practice can be difficult due to various factors: (i) patients may refuse to participate, (ii) patients may be too unwell, or (iii) there may be inadequate supervision from senior doctors. Simulated patient-doctor interactions using Objective Structured Clinical Examinations (OSCEs) may help bridge this gap. OSCEs are also used for assessing medical students, where a student-patient encounter is observed by an examiner and scored against a predefined matrix. However, providing these simulation opportunities for an entire cohort of students is resource intensive. In addition, simulated patients have limitations such as fatigue and variability in performance( 3 ). By contrast, an AI virtual patient (AIVP) may balance consistency and adaptability, and offer students the opportunity to practice OSCEs at their convenience to foster independent learning. This pilot project aims to create an AI virtual patient (AIVP) mimicking a patient presenting to an emergency department with a critical care pathology, and validate its ability to interact with a human user (a medical student). We will gather feedback from students on the fidelity and utility of the AIVP in performing this role and use this feedback to further train the AIVP. Methods We developed six AIVP scenarios, each representing a different critical care pathology and symptom presentation, simulating an emergency department interaction between a patient and a medical student within the OSCE framework. These pathologies were acute myocardial infarction (chest pain), subarachnoid haemorrhage (headache), neutropenic urosepsis (fever), musculoskeletal back pain, pulmonary embolism (shortness of breath), and ectopic pregnancy (abdominal pain). The AIVP was developed in-house using the Unity game engine to present a variety of 3D animated avatars where the students interact using their voice, with the AIVP responding in kind. These students used the web version available to them at any time during the study period. The main conversation loop involved converting speech to text (using OpenAI Whisper), generating responses (using Open AI ChatGPT 4o), and converting text back to speech (using OpenAI TTS). Additionally, a tutor AI (ChatGPT 4o) provides feedback after the conversation to students on how they could improve their structure and content. Phase 3 final year medical students were invited to participate in the AIVP project. After providing consent, the students participated in a formal OSCE conducted by a senior Emergency Physician and their performance was scored against expected academic standards. The students were then given access to the six AIVPs, each with different underlying pathologies and designed to train specific technical and non-technical skills. For example, one chatbot allowed students to practice taking a chest pain history from a compliant AIVP with a diagnosis of heart attack (acute coronary syndrome), while another enabled interaction with an agitated patient with back pain, upset about not receiving strong analgesia. Students were encouraged to interact with the AIVPs as many times as they wished over a trial period, allowing them to iterate and refine their technical and non-technical skills in medical history-taking. After 2 weeks, students were reassessed using a repeat OSCE to determine whether the AIVP had improved their performance. These assessments were formative and had no impact on students’ academic records. The students' scores before and after using the AIVP were compared using a non-parametric paired student t-test (Wilcoxon) to assess the effectiveness of the AIVP in improving history-taking skills. Additionally, students completed a mixed survey evaluating the utility of the AIVP as an educational tool, using a Likert scale and providing qualitative responses (Appendix 1). Thematic analysis was employed to analyse the data collected from the student survey. The data were reviewed, and initial themes were identified and refined. Three main themes—Feedback, Education, and Technology—were established, along with 14 sub-themes: Genetic, Timing, Quality, Accuracy, Student Communication with Patient, History Taking, Authenticity, Learning Tool, AIVP Behaviour, Design, Accessibility, and Latency. Further, each sub-theme was divided into positive and negative categories, resulting in a total of 28 distinct classifications. In total, 256 data points were identified across these classifications, acknowledging that individual comments or sentences could be relevant to multiple categories. The majority of qualitative comments were related to Education (44%), followed by Technology (33%) and Feedback (23%). This study was approved by the Research Ethics and Compliance Committee at the University of New South Wales Sydney (iRECS6086). Clinical Trial number: not applicable. Results A total of 21 students participated in the pilot study over two weeks, interacting with the AIVPs for a combined total 21.7 hours, with an average of 1.1 hours per user (highest user 2.4 hours, lowest user 0.2 hours). The median OSCE scores before and after using the AIVP were 63/100 (IQR: 53.5–70) and 70/100 (IQR: 63-73.5), respectively (p = 0.29) (Fig. 1 ). Twenty-one students participated in the student survey and expressed generally positive feedback across all items. On a Likert scale of 0 (strongly disagree) to 5 (strongly agree) there was particularly strong agreement that the AIVP was a valuable learning experience (mean 4.62, SD 0.65), provided a learning opportunity not met by other aspects of the course (mean 4.29, SD 0.93), and is a tool that should be used more widely in medical education (mean 4.52, SD 0.73) (Fig. 2 ). From the perspective of technical skills, students responded positively to statements that the AIVP ‘increased my confidence in my ability to take history from a patient’ (mean 4.24, SD 0.75) (Fig. 3 ), ‘improved my ability to discuss a likely diagnosis/ differential diagnosis with a patient’ (mean 4.48, SD 0.66), and ‘improved my ability to discuss next steps in assessment and management with a patient’ (mean 4.43, SD 0.58). Students also appeared to value the feedback provided by the AIVP at the end of their interaction, agreeing that the AIVP ‘provided me with useful feedback on my performance at the end of the interaction’ (mean 4.38, SD 0.84), and ‘delivered me with feedback in a style that was conductive to my learning’ (mean 4.33, SD 0.84) (Fig. 4 ). Figure 1 , Fig. 2 , Fig. 3 , Fig. 4 can be placed here. Students indicated that the AIVP did not perform so well in qualities of realism, with less strong agreement that the AIVP exhibited authentic character traits (mean 3.57, SD 0.85), and answered questions in a natural manner (mean 3.71, SD 0.98). This may be the reason that students were less inclined to agree that the AIVP improved communication skills - ‘improved my ability to communicate effectively with a patient’ (mean 3.81, SD 0.79) - than they did with its utility in improving their technical skills of history taking, diagnosis and management planning. Qualitative analysis of free-text feedback supported the quantitative data that the AIVP was a valuable learning tool for developing skills in history taking and providing timely, high-quality feedback on performance. However, it also highlighted that some students experienced technical difficulties using the AIVP. Some had issues with voice recognition, which may have been related to students who speak English as a second language. Additionally, some students noted a significant time lag (latency) in AIVP responses to their questions, which made interactions feel less realistic. Two students also experienced episodes of AIVP role reversal, where the AIVP switched between the role of patient and doctor during the interaction (which was resolved by logging out and back into the program). Two students discontinued using the program due to frequent technical glitches. Overall, the AIVP received positive feedback, with most reporting it to be a valuable learning experience, providing structured practice opportunities to develop clinical skills in a low-stress and iterative environment. Students highlighted its potential for enhancing learning through immediate feedback on their performance. However, improvements in the realism of interactions, technical performance, and user experience are necessary to maximise its utility in medical education. Discussion Our results suggest that it is feasible for students to access at home and on-demand experience with patient interaction through an AIVP. Unbounded from traditional working hours, the AIVP provides a tool to practice history-taking, clinical reasoning, synthesis of differential diagnosis, investigation and management planning, and patient communication. Previous research suggests that students with more exposure to patient assessment develop greater confidence in performing these tasks( 4 ). The AIVP allows these skills can be refined in a safe, less stressful environment, where interactions can be unlimited, and trial and error is encouraged and comes at no cost to patient comfort. Similar positive experience from students were found by another study looking at using a virtual patient with back pain for students to practice history taking( 5 ). Despite positive feedback, no significant improvement in pre- and post-AIVP OSCE assessment scores was observed in our students. This may be due to the short two-week period between pre- and post-AIVP OSCE assessments, which may not have allowed students enough time to internalise the feedback and format provided by the AIVP. Further, the pre- and post- AIVP OSCEs tested different clinical scenarios, requiring different clinical knowledge bases. A large focus of the AIVP experience is to improve transferable non-technical skills in patient interactions, and as such improvement in these areas may have been diluted by inadequate clinical knowledge. Personalised, constructive feedback from the AIVP at the end of the scenario is one of the program's greatest assets. Over eighty-five percent of students agreed or strongly agreed that the AIVP provided useful feedback on performance at the end of the interaction. Other studies have also confirmed the accuracy of AI feedback when compared with experienced clinicians( 6 ). In one study, a prospective trial involved medical students performing history-taking with a GPT-powered chatbot. The GPT model was found to be effective in providing structured feedback and had a 99% agreement with human raters( 6 ). This suggests that large language models like GPT can be valuable tools in medical education. Our findings support the careful integration of AI-driven feedback mechanisms into medical training. This pilot study has highlighted to researchers areas for improvement in the prototype AIVP mode. Technical issues were encountered and two students discontinued the program due to slow response times. The slow response time may be exacerbated by poor internet bandwidth and students will have to ensure a strong Wi-Fi connection when using the AIVP to get a more efficient patient-student interaction. In addition, the AIVP may struggle to replicate the complexity and spontaneity of real patient interactions, with some students commenting that the AIVP lacked emotional depth. Role reversal was an uncommon issue, reported by only 2 out of 21 students (and only then in a minority of interactions). These problems were resolved by logging out and logging back into the program. This finding aligns with other studies that used GPT-powered chatbots as virtual patients for medical history-taking practice( 5 , 7 ). One such study found 97% of responses were acceptable, although 16% of responses were missing, and some were inappropriate, leaving the role identity or providing illogical information( 7 ). This research group hypothesises that strengthening the AIVP prompts and providing more comprehensive background information may reduce the incidence of role-reversal. A limitation of this study is the small number of students in the pilot, which inhibits the generalisability of the results. However, this feasibility study provides valuable insights and will inform adjustments to improve the AIVP and the way it is made available in medical student education. Conclusions AIVP offers a positive learning experience for students that can be accessed at their convenience, from any location, in an environment where iterative interactions and trial-and-error is encouraged. AI feedback on student performance in these interactions appears to have great potential. However, adjustments are needed to improve realism, response time, feedback accuracy, and technical performance. Although no statistically significant improvement was observed in pre- and post-AIVP OSCE assessments, students reported finding the program helpful in improving their confidence in patient assessment. Abbreviations AIVP – Artificial Intelligence Virtual Patient OSCE – Objective Structured Clinical Education Declarations Ethics approval and consent to participate This study was approved by the Research Ethics and Compliance Committee at the University of New South Wales Sydney (iRECS6086). Clinical Trial number: not applicable. Informed consents to participate in the study have been obtained from participants. Documentary evidence of consent can be supplied if requested. Consent for publication Not applicable Availability of data and materials The data that support the findings of this study are available on request from the corresponding author, subject to ethical approval. The data are not publicly available due to privacy or ethical restrictions. Competing interest: The authors have no competing interests to report. Funding: This study has not received any external funding. Authors’ contribution TD, PD, GH provides the technical support to build the Artificial Intelligence. JX, LJ, BK provide the education support and tools for the evaluation of AIVP using the Likert Scale. JM and BSC develops the AIVP clinical scenarios. AT, BSC and JM develops the research idea and concept. BSC wrote the manuscript and all authors reviewed and revised the manuscript. Acknowledgements The authors would like to thank the School of Clinical Medicine, UNSW for providing the administrative support for the study. References Stamer T, Steinhauser J, Flagel K. Artificial Intelligence Supporting the Training of Communication Skills in the Education of Health Care Professions: Scoping Review. J Med Internet Res. 2023;25:e43311. Keifenheim KE, Teufel M, Ip J, Speiser N, Leehr EJ, Zipfel S, et al. Teaching history taking to medical students: a systematic review. BMC Med Educ. 2015;15:159. Hamilton A, Molzahn A, McLemore K. The Evolution From Standardized to Virtual Patients in Medical Education. Cureus. 2024;16(10):e71224. Reid KJ, Dodds AE, McColl GJ. Conducting patient assessments as a medical student: frequency, barriers, and facilitators. Teach Learn Med. 2014;26(2):153-9. Maicher KR, Stiff A, Scholl M, White M, Fosler-Lussier E, Schuler W, et al. Artificial intelligence in virtual standardized patients: Combining natural language understanding and rule based dialogue management to improve conversational fidelity. Med Teach. 2022:1-7. Holderried F, Stegemann-Philipps C, Herrmann-Werner A, Festl-Wietek T, Holderried M, Eickhoff C, et al. A Language Model-Powered Simulated Patient With Automated Feedback for History Taking: Prospective Study. JMIR Med Educ. 2024;10:e59213. Holderried F, Stegemann-Philipps C, Herschbach L, Moldt JA, Nevins A, Griewatz J, et al. A Generative Pretrained Transformer (GPT)-Powered Chatbot as a Simulated Patient to Practice History Taking: Prospective, Mixed Methods Study. JMIR Med Educ. 2024;10:e53961. Additional Declarations No competing interests reported. Supplementary Files Appendix1StudentQuestionnairev1.0.docx Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 16 May, 2026 Reviews received at journal 15 May, 2026 Reviewers agreed at journal 06 May, 2026 Reviews received at journal 27 May, 2025 Reviewers agreed at journal 30 Apr, 2025 Reviewers invited by journal 25 Apr, 2025 Editor assigned by journal 25 Apr, 2025 Submission checks completed at journal 24 Apr, 2025 First submitted to journal 24 Apr, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6272736","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":448916338,"identity":"976e826f-7d19-4a0b-9dda-8b833531645e","order_by":0,"name":"Betty S Chan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAxElEQVRIiWNgGAWjYDACCQYGZoYKKIeHeC1nSNbC2EaKFv7Zzc8eF847nCc/I4Hxwds2BnmDA4QsuXPM3HjmtsPFBjcSmA3ntjEYbiCkxUAiwUyad9vhxA3SCWzSvG0MjERoSf8mzTvncOL82Qnsv4Fa7InQkgO0peFwYsPtBDZmoJZEglokbuSUSc84lp644f7DZsk55ySSZxLSwj8jfZt0QY114vyewwc/vCmzse0jpAUJMDYwgKNpFIyCUTAKRgHlAAC5Bj+YCuEP+gAAAABJRU5ErkJggg==","orcid":"","institution":"UNSW Sydney","correspondingAuthor":true,"prefix":"","firstName":"Betty","middleName":"S","lastName":"Chan","suffix":""},{"id":448916339,"identity":"09d06976-bce3-4818-b83c-3da84de87e1b","order_by":1,"name":"Timothy Dodds","email":"","orcid":"","institution":"UNSW Sydney","correspondingAuthor":false,"prefix":"","firstName":"Timothy","middleName":"","lastName":"Dodds","suffix":""},{"id":448916340,"identity":"017ae20a-3984-4946-9e9f-c92bc92aded1","order_by":2,"name":"Jen Xiang","email":"","orcid":"","institution":"UNSW Sydney","correspondingAuthor":false,"prefix":"","firstName":"Jen","middleName":"","lastName":"Xiang","suffix":""},{"id":448916341,"identity":"eda4bc8f-e7a9-4f76-a943-84977870d5ae","order_by":3,"name":"Lucy Jo","email":"","orcid":"","institution":"UNSW Sydney","correspondingAuthor":false,"prefix":"","firstName":"Lucy","middleName":"","lastName":"Jo","suffix":""},{"id":448916342,"identity":"03219f2c-4802-45c0-99f5-644cf2e8e57a","order_by":4,"name":"Paul Dyer","email":"","orcid":"","institution":"UNSW Sydney","correspondingAuthor":false,"prefix":"","firstName":"Paul","middleName":"","lastName":"Dyer","suffix":""},{"id":448916343,"identity":"2227690f-7e65-4578-989a-fe09e0dc3aae","order_by":5,"name":"Bhavani Kannan","email":"","orcid":"","institution":"UNSW Sydney","correspondingAuthor":false,"prefix":"","firstName":"Bhavani","middleName":"","lastName":"Kannan","suffix":""},{"id":448916344,"identity":"18b09688-90e3-4f5c-b7e3-77a99ede52e1","order_by":6,"name":"Graham Hannah","email":"","orcid":"","institution":"UNSW Sydney","correspondingAuthor":false,"prefix":"","firstName":"Graham","middleName":"","lastName":"Hannah","suffix":""},{"id":448916345,"identity":"c144f371-731c-4d20-ad69-017fde3af648","order_by":7,"name":"Adrienne Torda","email":"","orcid":"","institution":"UNSW Sydney","correspondingAuthor":false,"prefix":"","firstName":"Adrienne","middleName":"","lastName":"Torda","suffix":""},{"id":448916346,"identity":"75ee7ad9-6525-4b83-b8d7-a04ff6a7fff3","order_by":8,"name":"Jack Marjot","email":"","orcid":"","institution":"UNSW Sydney","correspondingAuthor":false,"prefix":"","firstName":"Jack","middleName":"","lastName":"Marjot","suffix":""}],"badges":[],"createdAt":"2025-03-20 23:08:15","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6272736/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6272736/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":82204582,"identity":"16811d68-3cad-4ea0-a17e-e6dafb4fd1a6","added_by":"auto","created_at":"2025-05-07 16:55:25","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":39242,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA Violin plot of the pre- and post-AIVP intervention OSCE assessment results (n=21).\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe black line represents the median OSCE score, with each dot representing 1 participant.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6272736/v1/6d41f0d563f5ab3c6cbfd722.png"},{"id":82204856,"identity":"6a5f6227-113d-4eb1-9167-2f5589531a83","added_by":"auto","created_at":"2025-05-07 17:03:25","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":176488,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStudent perceptions of learning experience using AI Virtual Patient.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6272736/v1/7f071986036f95b97a0773b6.png"},{"id":82204585,"identity":"e552a39d-6d40-4cf3-b8f4-d4cb5b4974ea","added_by":"auto","created_at":"2025-05-07 16:55:25","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":478583,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStudent perceptions of skills development when using AI Virtual Patient.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6272736/v1/38a0335688ed0608340ffd8c.png"},{"id":82205386,"identity":"2eae2057-0fe8-4088-9f05-ad9ca4efb9e3","added_by":"auto","created_at":"2025-05-07 17:19:25","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":248048,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStudent perceptions of quality of interaction when using AI Virtual Patient.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-6272736/v1/6dc8c6371792aa5b3db071db.png"},{"id":82203505,"identity":"adeadcda-eeb8-465d-a06c-c3a398e4a030","added_by":"auto","created_at":"2025-05-07 16:47:25","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":356383,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6272736/v1/2bd300ad-eeb4-456a-b91f-81aa30560f4a.pdf"},{"id":82204583,"identity":"e3121050-7483-4d58-bfd2-cd5f6f149ee0","added_by":"auto","created_at":"2025-05-07 16:55:25","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":62426,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix1StudentQuestionnairev1.0.docx","url":"https://assets-eu.researchsquare.com/files/rs-6272736/v1/4c156cce3b1fad1fedc2866f.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Artificial Intelligence Virtual Patient: A proof of concept study","fulltext":[{"header":"Introduction","content":"\u003cp\u003eArtificial Intelligence (AI) has the potential to recreate high-fidelity human interactions with users. Early examples in academic literature demonstrate AI being trained to assume the role of a patient in a medical setting(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). This is a rapidly evolving technology, and its application in medical education is still in its infancy.\u003c/p\u003e \u003cp\u003eIn medicine, taking a medical history is a fundamental skill for any medical professional. It is estimated that 80% of diagnoses can be made from history alone(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). This process involves asking the patient a series of questions to gather sufficient, holistic information to make a provisional diagnosis. Diagnostic reasoning then follows, along with an explanation to the patient about the next steps in assessment, investigation, and treatment. However, finding patients in hospitals for history-taking practice can be difficult due to various factors: (i) patients may refuse to participate, (ii) patients may be too unwell, or (iii) there may be inadequate supervision from senior doctors. Simulated patient-doctor interactions using Objective Structured Clinical Examinations (OSCEs) may help bridge this gap. OSCEs are also used for assessing medical students, where a student-patient encounter is observed by an examiner and scored against a predefined matrix. However, providing these simulation opportunities for an entire cohort of students is resource intensive. In addition, simulated patients have limitations such as fatigue and variability in performance(\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e). By contrast, an AI virtual patient (AIVP) may balance consistency and adaptability, and offer students the opportunity to practice OSCEs at their convenience to foster independent learning.\u003c/p\u003e \u003cp\u003eThis pilot project aims to create an AI virtual patient (AIVP) mimicking a patient presenting to an emergency department with a critical care pathology, and validate its ability to interact with a human user (a medical student). We will gather feedback from students on the fidelity and utility of the AIVP in performing this role and use this feedback to further train the AIVP.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eWe developed six AIVP scenarios, each representing a different critical care pathology and symptom presentation, simulating an emergency department interaction between a patient and a medical student within the OSCE framework. These pathologies were acute myocardial infarction (chest pain), subarachnoid haemorrhage (headache), neutropenic urosepsis (fever), musculoskeletal back pain, pulmonary embolism (shortness of breath), and ectopic pregnancy (abdominal pain). The AIVP was developed in-house using the Unity game engine to present a variety of 3D animated avatars where the students interact using their voice, with the AIVP responding in kind. These students used the web version available to them at any time during the study period. The main conversation loop involved converting speech to text (using OpenAI Whisper), generating responses (using Open AI ChatGPT 4o), and converting text back to speech (using OpenAI TTS). Additionally, a tutor AI (ChatGPT 4o) provides feedback after the conversation to students on how they could improve their structure and content.\u003c/p\u003e \u003cp\u003ePhase 3 final year medical students were invited to participate in the AIVP project. After providing consent, the students participated in a formal OSCE conducted by a senior Emergency Physician and their performance was scored against expected academic standards. The students were then given access to the six AIVPs, each with different underlying pathologies and designed to train specific technical and non-technical skills. For example, one chatbot allowed students to practice taking a chest pain history from a compliant AIVP with a diagnosis of heart attack (acute coronary syndrome), while another enabled interaction with an agitated patient with back pain, upset about not receiving strong analgesia. Students were encouraged to interact with the AIVPs as many times as they wished over a trial period, allowing them to iterate and refine their technical and non-technical skills in medical history-taking. After 2 weeks, students were reassessed using a repeat OSCE to determine whether the AIVP had improved their performance. These assessments were formative and had no impact on students\u0026rsquo; academic records. The students' scores before and after using the AIVP were compared using a non-parametric paired student t-test (Wilcoxon) to assess the effectiveness of the AIVP in improving history-taking skills. Additionally, students completed a mixed survey evaluating the utility of the AIVP as an educational tool, using a Likert scale and providing qualitative responses (Appendix 1).\u003c/p\u003e \u003cp\u003eThematic analysis was employed to analyse the data collected from the student survey. The data were reviewed, and initial themes were identified and refined. Three main themes\u0026mdash;Feedback, Education, and Technology\u0026mdash;were established, along with 14 sub-themes: Genetic, Timing, Quality, Accuracy, Student Communication with Patient, History Taking, Authenticity, Learning Tool, AIVP Behaviour, Design, Accessibility, and Latency. Further, each sub-theme was divided into positive and negative categories, resulting in a total of 28 distinct classifications. In total, 256 data points were identified across these classifications, acknowledging that individual comments or sentences could be relevant to multiple categories. The majority of qualitative comments were related to Education (44%), followed by Technology (33%) and Feedback (23%).\u003c/p\u003e \u003cp\u003e This study was approved by the Research Ethics and Compliance Committee at the University of New South Wales Sydney (iRECS6086). Clinical Trial number: not applicable.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eA total of 21 students participated in the pilot study over two weeks, interacting with the AIVPs for a combined total 21.7 hours, with an average of 1.1 hours per user (highest user 2.4 hours, lowest user 0.2 hours).\u003c/p\u003e \u003cp\u003eThe median OSCE scores before and after using the AIVP were 63/100 (IQR: 53.5\u0026ndash;70) and 70/100 (IQR: 63-73.5), respectively (p\u0026thinsp;=\u0026thinsp;0.29) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Twenty-one students participated in the student survey and expressed generally positive feedback across all items. On a Likert scale of 0 (strongly disagree) to 5 (strongly agree) there was particularly strong agreement that the AIVP was a valuable learning experience (mean 4.62, SD 0.65), provided a learning opportunity not met by other aspects of the course (mean 4.29, SD 0.93), and is a tool that should be used more widely in medical education (mean 4.52, SD 0.73) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). From the perspective of technical skills, students responded positively to statements that the AIVP \u0026lsquo;increased my confidence in my ability to take history from a patient\u0026rsquo; (mean 4.24, SD 0.75) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e), \u0026lsquo;improved my ability to discuss a likely diagnosis/ differential diagnosis with a patient\u0026rsquo; (mean 4.48, SD 0.66), and \u0026lsquo;improved my ability to discuss next steps in assessment and management with a patient\u0026rsquo; (mean 4.43, SD 0.58). Students also appeared to value the feedback provided by the AIVP at the end of their interaction, agreeing that the AIVP \u0026lsquo;provided me with useful feedback on my performance at the end of the interaction\u0026rsquo; (mean 4.38, SD 0.84), and \u0026lsquo;delivered me with feedback in a style that was conductive to my learning\u0026rsquo; (mean 4.33, SD 0.84) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e \u003cem\u003ecan be placed here.\u003c/em\u003e\u003c/p\u003e \u003cp\u003eStudents indicated that the AIVP did not perform so well in qualities of realism, with less strong agreement that the AIVP exhibited authentic character traits (mean 3.57, SD 0.85), and answered questions in a natural manner (mean 3.71, SD 0.98). This may be the reason that students were less inclined to agree that the AIVP improved communication skills - \u0026lsquo;improved my ability to communicate effectively with a patient\u0026rsquo; (mean 3.81, SD 0.79) - than they did with its utility in improving their technical skills of history taking, diagnosis and management planning.\u003c/p\u003e \u003cp\u003eQualitative analysis of free-text feedback supported the quantitative data that the AIVP was a valuable learning tool for developing skills in history taking and providing timely, high-quality feedback on performance. However, it also highlighted that some students experienced technical difficulties using the AIVP. Some had issues with voice recognition, which may have been related to students who speak English as a second language. Additionally, some students noted a significant time lag (latency) in AIVP responses to their questions, which made interactions feel less realistic. Two students also experienced episodes of AIVP role reversal, where the AIVP switched between the role of patient and doctor during the interaction (which was resolved by logging out and back into the program). Two students discontinued using the program due to frequent technical glitches.\u003c/p\u003e \u003cp\u003eOverall, the AIVP received positive feedback, with most reporting it to be a valuable learning experience, providing structured practice opportunities to develop clinical skills in a low-stress and iterative environment. Students highlighted its potential for enhancing learning through immediate feedback on their performance. However, improvements in the realism of interactions, technical performance, and user experience are necessary to maximise its utility in medical education.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eOur results suggest that it is feasible for students to access at home and on-demand experience with patient interaction through an AIVP. Unbounded from traditional working hours, the AIVP provides a tool to practice history-taking, clinical reasoning, synthesis of differential diagnosis, investigation and management planning, and patient communication. Previous research suggests that students with more exposure to patient assessment develop greater confidence in performing these tasks(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). The AIVP allows these skills can be refined in a safe, less stressful environment, where interactions can be unlimited, and trial and error is encouraged and comes at no cost to patient comfort. Similar positive experience from students were found by another study looking at using a virtual patient with back pain for students to practice history taking(\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDespite positive feedback, no significant improvement in pre- and post-AIVP OSCE assessment scores was observed in our students. This may be due to the short two-week period between pre- and post-AIVP OSCE assessments, which may not have allowed students enough time to internalise the feedback and format provided by the AIVP. Further, the pre- and post- AIVP OSCEs tested different clinical scenarios, requiring different clinical knowledge bases. A large focus of the AIVP experience is to improve transferable non-technical skills in patient interactions, and as such improvement in these areas may have been diluted by inadequate clinical knowledge.\u003c/p\u003e \u003cp\u003ePersonalised, constructive feedback from the AIVP at the end of the scenario is one of the program's greatest assets. Over eighty-five percent of students agreed or strongly agreed that the AIVP provided useful feedback on performance at the end of the interaction. Other studies have also confirmed the accuracy of AI feedback when compared with experienced clinicians(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). In one study, a prospective trial involved medical students performing history-taking with a GPT-powered chatbot. The GPT model was found to be effective in providing structured feedback and had a 99% agreement with human raters(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). This suggests that large language models like GPT can be valuable tools in medical education. Our findings support the careful integration of AI-driven feedback mechanisms into medical training.\u003c/p\u003e \u003cp\u003eThis pilot study has highlighted to researchers areas for improvement in the prototype AIVP mode. Technical issues were encountered and two students discontinued the program due to slow response times. The slow response time may be exacerbated by poor internet bandwidth and students will have to ensure a strong Wi-Fi connection when using the AIVP to get a more efficient patient-student interaction. In addition, the AIVP may struggle to replicate the complexity and spontaneity of real patient interactions, with some students commenting that the AIVP lacked emotional depth.\u003c/p\u003e \u003cp\u003eRole reversal was an uncommon issue, reported by only 2 out of 21 students (and only then in a minority of interactions). These problems were resolved by logging out and logging back into the program. This finding aligns with other studies that used GPT-powered chatbots as virtual patients for medical history-taking practice(\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e). One such study found 97% of responses were acceptable, although 16% of responses were missing, and some were inappropriate, leaving the role identity or providing illogical information(\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e). This research group hypothesises that strengthening the AIVP prompts and providing more comprehensive background information may reduce the incidence of role-reversal.\u003c/p\u003e \u003cp\u003eA limitation of this study is the small number of students in the pilot, which inhibits the generalisability of the results. However, this feasibility study provides valuable insights and will inform adjustments to improve the AIVP and the way it is made available in medical student education.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eAIVP offers a positive learning experience for students that can be accessed at their convenience, from any location, in an environment where iterative interactions and trial-and-error is encouraged. AI feedback on student performance in these interactions appears to have great potential. However, adjustments are needed to improve realism, response time, feedback accuracy, and technical performance. Although no statistically significant improvement was observed in pre- and post-AIVP OSCE assessments, students reported finding the program helpful in improving their confidence in patient assessment.\u003c/p\u003e "},{"header":"Abbreviations","content":"\u003cp\u003eAIVP \u0026ndash; Artificial Intelligence Virtual Patient\u003c/p\u003e\n\u003cp\u003eOSCE \u003cstrong\u003e\u0026ndash;\u0026nbsp;\u003c/strong\u003eObjective Structured Clinical Education\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was approved by the Research Ethics and Compliance Committee at the University of New South Wales Sydney (iRECS6086). \u0026nbsp;Clinical Trial number: not applicable.\u003c/p\u003e\n\u003cp\u003eInformed consents to participate in the study have been obtained from participants. \u0026nbsp;Documentary evidence of consent can be supplied if requested.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the findings of this study are available on request from the corresponding author, subject to ethical approval. \u0026nbsp;The data are not publicly available due to privacy or ethical restrictions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interest:\u0026nbsp;\u003c/strong\u003eThe authors have no competing interests to report.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u0026nbsp;\u003c/strong\u003eThis study has not received any external funding.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; contribution\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTD, PD, GH provides the technical support to build the Artificial Intelligence. \u0026nbsp;JX, LJ, BK provide the education support and tools for the evaluation of AIVP using the Likert Scale. \u0026nbsp;JM and BSC develops the AIVP clinical scenarios. \u0026nbsp;AT, BSC and JM develops the research idea and concept. \u0026nbsp;BSC wrote the manuscript and all authors reviewed and revised the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors would like to thank the School of Clinical Medicine, UNSW for providing the administrative support for the study.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eStamer T, Steinhauser J, Flagel K. Artificial Intelligence Supporting the Training of Communication Skills in the Education of Health Care Professions: Scoping Review. J Med Internet Res. 2023;25:e43311.\u003c/li\u003e\n\u003cli\u003eKeifenheim KE, Teufel M, Ip J, Speiser N, Leehr EJ, Zipfel S, et al. Teaching history taking to medical students: a systematic review. BMC Med Educ. 2015;15:159.\u003c/li\u003e\n\u003cli\u003eHamilton A, Molzahn A, McLemore K. The Evolution From Standardized to Virtual Patients in Medical Education. Cureus. 2024;16(10):e71224.\u003c/li\u003e\n\u003cli\u003eReid KJ, Dodds AE, McColl GJ. Conducting patient assessments as a medical student: frequency, barriers, and facilitators. Teach Learn Med. 2014;26(2):153-9.\u003c/li\u003e\n\u003cli\u003eMaicher KR, Stiff A, Scholl M, White M, Fosler-Lussier E, Schuler W, et al. Artificial intelligence in virtual standardized patients: Combining natural language understanding and rule based dialogue management to improve conversational fidelity. Med Teach. 2022:1-7.\u003c/li\u003e\n\u003cli\u003eHolderried F, Stegemann-Philipps C, Herrmann-Werner A, Festl-Wietek T, Holderried M, Eickhoff C, et al. A Language Model-Powered Simulated Patient With Automated Feedback for History Taking: Prospective Study. JMIR Med Educ. 2024;10:e59213.\u003c/li\u003e\n\u003cli\u003eHolderried F, Stegemann-Philipps C, Herschbach L, Moldt JA, Nevins A, Griewatz J, et al. A Generative Pretrained Transformer (GPT)-Powered Chatbot as a Simulated Patient to Practice History Taking: Prospective, Mixed Methods Study. JMIR Med Educ. 2024;10:e53961.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-education","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"meed","sideBox":"Learn more about [BMC Medical Education](http://bmcmededuc.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/meed/default.aspx","title":"BMC Medical Education","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6272736/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6272736/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eArtificial Intelligence(AI) is advancing, but its role in simulating detailed patient-doctor interactions in the style of Objective Structured Clinical Examinations(OSCEs) is emerging. This study's goal was to create and validate an AI virtual patient(AIVP) that could interact with medical students, mimic a patient with a medical issue, and provide students feedback on their performance.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eSix AIVP were developed to simulate OSCE scenarios for common emergency department presentations. The simulations were created using the Unity game engine, featuring a conversation loop that includes speech-to-text conversion (OpenAI Whisper), response generation(Open AI ChatGPT 4o), and speech generation (OpenAI TTS). A tutor AI(ChatGPT 4o) then generates feedback after the conversation to help students improve their responses. Final-year medical students were given the opportunity to interact with the AIVPs and participated in pre- and post-AIVP OSCE assessments to evaluate the AIVP's effect on performance, with Wilcoxon paired t-tests used for analysis. Students completed Likert Scales and surveys on the AIVP's educational value and technical issues.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eTwenty-one students participated over two weeks for a total of 21.7 hours, averaging 1.1 hour per user. The median OSCE scores improved from 63/100 (IQR: 53.5\u0026ndash;70) to 70/100 (IQR: 63-73.5) (p\u0026thinsp;=\u0026thinsp;0.29). On a Likert scale of 0 (strongly disagree) to 5 (strongly agree) there was strong agreement that the AIVP was a valuable learning experience(mean 4.62, SD 0.65). Students valued the feedback provided by the AIVP at the end of their interaction on their performance(mean 4.38, SD 0.84), Technical issues like voice recognition problems, latency in AIVP interaction, and occasional role reversals were reported.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThis is a novel tool for developing history-taking skills and OSCE performance. Students found their interactions with the AIVP, and the feedback it provided on their performance, to be a valuable learning experience. However, technical factors and AIVP realism need further development.\u003c/p\u003e","manuscriptTitle":"Artificial Intelligence Virtual Patient: A proof of concept study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-07 16:47:20","doi":"10.21203/rs.3.rs-6272736/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-05-16T05:14:11+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-15T21:35:55+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"59723997267862359628642540065666989507","date":"2026-05-06T06:15:52+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-05-27T22:02:49+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"125119481349679677534824733895700649381","date":"2025-04-30T19:16:16+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-04-25T19:00:01+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-04-25T10:13:25+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-04-25T01:32:37+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Education","date":"2025-04-25T01:31:28+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-education","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"meed","sideBox":"Learn more about [BMC Medical Education](http://bmcmededuc.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/meed/default.aspx","title":"BMC Medical Education","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"14025aa2-5866-41a1-99c5-8bfd7e4c258d","owner":[],"postedDate":"May 7th, 2025","published":true,"recentEditorialEvents":[{"type":"decision","content":"Revision requested","date":"2026-05-16T05:14:11+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-15T21:35:55+00:00","index":191,"fulltext":""},{"type":"reviewerAgreed","content":"59723997267862359628642540065666989507","date":"2026-05-06T06:15:52+00:00","index":190,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-05-16T05:24:40+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-07 16:47:20","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6272736","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6272736","identity":"rs-6272736","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}