Teaching communication skills to anesthesiology students using ChatGPT as a simulated patient and the SEGUE framework

Teaching communication skills to anesthesiology students using ChatGPT as a simulated patient and the SEGUE framework | 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 Teaching communication skills to anesthesiology students using ChatGPT as a simulated patient and the SEGUE framework Peng Zhang, Yu Liu, Jing Zhuang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8898929/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract Background In this randomized controlled trial, we aimed to evaluate the efficacy of a one-month training program using ChatGPT-simulated patient and automated SEGUE framework feedback in improving the preoperative communication skills of anesthesiology students. Methods Thirty-four anesthesiology students were randomly allocated to either a ChatGPT group (n = 17) or a control group (n = 17). The ChatGPT group underwent a structured training program, conducting at least three simulated preoperative assessments per week with ChatGPT-4.0, which provided immediate SEGUE scores and feedback. The control group received only baseline and endpoint assessments with ChatGPT-4.0 as well, but without SEGUE feedback. The primary outcome was the change in SEGUE scores, which range from 0 to 25. Results The ChatGPT group showed a faster improvement trend in SEGUE scores than the control group, suggesting a significant Time × Group interaction effect (F = 44.51, p < 0.001). The ChatGPT group obtained higher SEGUE scores (from 15.76 ± 2.86 to 22.29 ± 1.36) compared to the control group (from 15.65 ± 2.81 to 16.88 ± 2.47) after one month (p = 0.01). Satisfaction survey indicated that 94% of the ChatGPT group participants found the method acceptable (47% satisfied, 30% completely satisfied). Conclusions A ChatGPT-assisted training program significantly enhanced the preoperative communication skills of anesthesiology students and was well-received. This approach represents a promising, scalable tool for complementing traditional communication skills training in medical education. Trial registration: This study was registered in the Chinese Clinical Trial Registry (ChiCTR2500106474) on July 24, 2025. ChatGPT medical education AI medical research Figures Figure 1 Introduction Anesthesiologists, despite often being perceived as physicians who operate behind the scenes, still require robust patient communication skills, particularly during the preoperative visit process. Preoperative anesthesiologist-patient interactions are frequently complicated by several factors, including time constraints, patient anxiety, a noisy bedside environment, and absence of a pre-existing relationship. Traditional medical education primarily focuses on teaching students basic medical theory, but tends to neglect the development of doctor-patient communication skills [ 1 ]. This situation often results in medical students not feeling confident enough to communicate with patients in real clinical settings. Standardized patients and peer role-playing are two fundamental simulation-based medical education methods for training technical (procedural) skills and communication skills in medical students [ 2 – 4 ]. Simulated patients, who are professional actors or trained individuals, are used to teach communication skills by delivering standardized and customizable scenarios and providing feedback [ 5 , 6 ]. While they offer a controlled learning environment, their use can be expensive, time-consuming, and occasionally inconsistent despite training [ 2 ]. Role-play among medical students is a cost-effective way to practice communication skills in a simulated setting [ 7 ]. However, it may feel artificial and encounter resistance from trainees who are reluctant to take part in role-play with colleagues [ 3 ]. ChatGPT, developed by OpenAI, is capable of interacting with individual users and generating human-like responses to inputs in a conversational style. This technology has been documented in the medical education. Moreover, some authors have reported its novel use as a role-playing device to generate human-like responses in fictional but realistic clinical scenarios for educational purposes. Webb et al. have utilized ChatGPT-3.5 as an educational tool to train emergency physicians in the breaking bad news to patients [ 8 ]. Wang et al. have demonstrated that ChatGPT can effectively simulate a standardized patient by inputting appropriate prompts, showing its potential to improve medical training [ 9 ]. In this study, we aimed to use ChatGPT-4.0 to design a training scenario in which anesthesiology students would engage in role-play tasks with a simulated patient to practice preoperative assessments. Additionally, we also aimed to evaluate whether the communication skills of the students improved following the one month training. Methods Ethics consideration This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Clinical and Biomedical Trials at Sichuan Provincial People's Hospital (Approval No. 2025 − 389). Each participant provided written informed consent. This trial was registered prior to patient enrollment with the Chinese Clinical Trial Registry (Registration Number: ChiCTR2500106474). The principal investigator for the registration was Dr. Jing Zhuang, and the date of registration was 2025-07-24. No patient or public was involved in the design, conduct, or reporting of this study. This randomized controlled trial is reported in accordance with the CONSORT (Consolidated Standards of Reporting Trials) 2025 guidelines (supplementary file 1) [ 10 ]. Study participants This parallel-group randomized trial was conducted at Sichuan Provincial People's Hospital, Chengdu, Sichuan, China. No changes were made to the trial protocol after commencement. The inclusion criteria were as follows: anesthesiology students (including postgraduate students and residents) who are scheduled for a clinical rotation in the Department of Anesthesiology, Sichuan Provincial People's Hospital, from July 2025 to March 2026. The exclusion criteria were as follows: (1) individuals unable to operate electronic devices (e.g., computers, tablets, or smartphones); and (2) individuals without internet access. The allocation sequence was generated using a computer software (SPSS, version 26.0) by an independent researcher using simple randomization, with no stratification or blocking. An independent researcher sealed the sequence in consecutively numbered, opaque, sealed envelopes to conceal the allocation. The investigators responsible for participant enrollment and intervention were kept unaware of the sequence. This study recruited a cohort of 34 participants. All participants have completed both fundamental medical theory courses and clinical practice components at the undergraduate level. Participants were not blinded to their group allocation due to the visible difference in receiving SEGUE feedback. ChatGPT-4 simulated as a standardized patent The initial input prompt was designed to establish a medical scenario involving anesthesiology, with ChatGPT assigned the role of the patient. This prompt was adapted from a study by Wang et al. [ 9 ], which used ChatGPT as a standardized patient for history-taking tasks. The effectiveness of this prompt have been validated based on its anthropomorphism, clinical accuracy, and adaptability [ 9 ]. Acting as the anesthesiologist, the trainees provided additional inputs in response to ChatGPT's outputs to simulate a realistic and natural conversation during the preoperative visit process. The input and output responses were recorded using screen capture software. The input prompts used to design ChatGPT-4 as the simulated patient are detailed in supplementary file 2. SEGUE framework The SEGUE framework, developed by Professor Makoul at the Feinberg School of Medicine, Northwestern University, is designed to guide medical students in communicating effectively with patients [ 11 ]. SEGUE is an acronym for its five core components: set the stage, elicit information, give information, understand the patient’s perspective and end the encounter [ 11 ]. The SEGUE framework employs a binary scoring system, where each item was rated as either "Yes" (score = 1) or "No" (score = 0) [ 11 ]. The total score for each encounter was calculated by summing the scores for each item, with higher scores indicating better provider-patient communication skills [ 11 ]. We used a Chinese version of the SEGUE to measure the communication performance of the trainees [ 12 ]. Overall satisfaction evaluation Overall satisfaction was measured using a 5-point Likert scale: 1 = completely dissatisfied, 2 = dissatisfied, 3 = partially satisfied, 4 = satisfied, and 5 = completely satisfied. Higher total scores indicate greater respondent satisfaction [ 13 ]. Study procedure Participants in the ChatGPT-assisted group underwent a structured training protocol, beginning with an initial ChatGPT-simulated preoperative assessment, after which ChatGPT-4 automatically generated a SEGUE score and provided structured feedback based on the SEGUE framework. Over the subsequent month, they were required to independently complete this simulated assessment at least three times per week, each followed by real-time SEGUE feedback. In contrast, control group participants were assessed using the same ChatGPT system at baseline and the study endpoint but received no SEGUE feedback or additional training. Participants were not blinded to their group allocation. All SEGUE scoring was performed automatically by ChatGPT-4 using de-identified conversation transcripts. Endpoint assessments employed parallel case scenarios. Finally, an overall satisfaction evaluation (using a 5-point Likert scale) was administered by a research assistant who was blinded to the group assignments. Outcomes The primary outcome was the change in SEGUE scores from baseline to one-month follow-up, measured as the mean ± standard deviation for each group. The secondary outcome was overall satisfaction with the ChatGPT-assisted training method, assessed using a 5-point Likert scale at the end of the study period. No harms or adverse events were assessed, as the intervention was educational. Outcome assessment was performed by a research assistant who was blinded to group assignments. The data analyst was also blinded to group allocation. Statistical analysis Statistical analyses were performed using SPSS (Version 26.0; IBM Corp., Armonk, NY, USA). The sample size was determined using G*Power (Version 3.1.9.7; Heinrich-Heine-Universität Düsseldorf, Düsseldorf, Germany). Based on an anticipated medium effect size (f = 0.25), an α level of 0.05, and a desired power of 0.8, the calculated total sample size was 34 participants. No interim analyses were planned or conducted. Continuous variables are presented as mean ± standard deviation (SD), and categorical data are reported as numbers (%). Baseline differences between the two groups in terms of sex, age, and education level were assessed using independent samples t-tests. A repeated-measures analysis of variance (ANOVA) was employed to compare the SEGUE scores between groups across two time points (baseline and one-month follow-up). All analyses were performed on an intention-to-treat basis, including all randomized participants. A p-value of less than 0.05 was considered statistically significant. Results Participants were recruited from August 2025 to January 2026, with the one-month follow-up completed by February 2026. A total of 98 anesthesiology students scheduled for their clinical rotation in the Department of Anesthesiology were assessed for eligibility and met the inclusion criteria. Of these, 64 were excluded: 58 because they were not assigned to Dr. Zhang for clinical supervision during the study period, 2 because they lacked electronic devices or internet access, and 4 because they declined to participate. Finally, 34 students were enrolled and randomized to the ChatGPT group (n = 17) or the control group (n = 17). All 34 participants received the intended intervention and were included in the primary outcome analysis. There were no losses to follow-up or exclusions after randomization. The participant flow diagram is shown in Fig. 1 . The ChatGPT group comprised five male and 12 female participants (mean age = 24.06 ± 0.83 years), while the control group consisted of eight male and nine female participants (mean age = 24.41 ± 0.71 years). There was no significant difference in sex (p = 0.481) and age (p = 0.192) between the two groups. Participants in the ChatGPT group completed a mean of 13.18 ± 1.38 (range 12 to 16) simulated preoperative assessment training during the one-month period. No participants in the control group reported engaging in any additional communication skills training during the study period. Main effect of time: The within-subjects effect test indicated a significant time effect on SEGUE scores (F = 95.74, p < 0.001), demonstrating statistically significant differences in SEGUE scores across different time points. Time × Group interaction: The interaction between time and group was significant (F = 44.51, p < 0.001), suggesting different trends of SEGUE score changes over time between groups. Specifically, the ChatGPT group showed significantly faster score improvement than the control group. Between-group main effect: The between-subjects effect test revealed statistically significant differences in overall SEGUE scores between groups (F = 13.93, p = 0.01), indicating that the intervention effects on SEGUE scores differed between groups, with the ChatGPT group demonstrating superior intervention effectiveness compared to the control group (Table 1 ). The baseline SEGUE scores were comparable between the two groups (p = 0.904). Both groups exhibited significant improvements in SEGUE scores compared to their baseline values (p < 0.001 for the ChatGPT group and p = 0.035 for the control group). Notably, the SEGUE scores of the ChatGPT group were significantly higher than those of the control group after three months of training (22.29 ± 1.36 vs. 16.88 ± 2.47, p < 0.001) (Table 1 ). Overall satisfaction was assessed only in the ChatGPT group. Analysis of the survey data revealed favorable acceptance of the ChatGPT-assisted teaching methodology in the ChatGPT group: no participants (0%) expressed complete dissatisfaction, one participant (6%) expressed dissatisfaction, three participants (17%) expressed partial satisfaction, eight participants (47%) expressed satisfaction, and five participants (30%) expressed complete satisfaction (Table 2 ). Table 1 Results of repeated measures ANOVA of SEGUE scores SEGUE scores before training ChatGPT group Control group F P 15.76 ± 2.86 15.65 ± 2.81 0.15 0.904 SEGUE scores after training 22.29 ± 1.36 16.88 ± 2.47 62.58 < 0.001 F 135.40 4.85 P < 0.001 0.035 Between-group main effect F = 13.93 P = 0.01 Main effect of time F = 95.74 P < 0.001 Time × Group interaction F = 44.51 P < 0.001 Table 2 Residents’ satisfaction evaluation of ChatGPT training Satisfaction evaluation Number of respondents Percentage Completely dissatisfied 0 —— Dissatisfied 1 6% Partially satisfied 3 17% Satisfied 8 47% Completely satisfied 5 30% Discussion In this study, we investigated a convenient and interactive method to teach communication skills to anesthesiology students. We found that a one-month training program with a combination teaching method, using a ChatGPT-simulated patient and the SEGUE framework, significantly improved the preoperative communication skills of anesthesiology trainees. Furthermore, the overall satisfaction survey indicated that the majority of participants found this AI-assisted teaching method acceptable and beneficial. These findings highlight the potential of large language models like ChatGPT as a flexible and effective tool for simulating clinical interactions and enhancing communication skills training in medical education. Our findings reinforce a small but growing body of literature exploring the use of ChatGPT in clinical role-playing [ 14 ]. For instance, Webb et al. and Wang et al. have reported its successful application in simulating patients for training in breaking bad news and in history-taking tasks, respectively [ 8 , 9 ]. The present study extends these findings by applying this technology specifically to the anesthesiology preoperative assessment—a context characterized by time pressure and the need to rapidly build trust and gather complex information [ 15 ]. Compared to the control group, our results showed that the ChatGPT group's SEGUE scores increased more rapidly (as reflected in the significant Time × Group interaction effect) and reached higher levels. This outcome demonstrates that the iterative practice (three times per week) with immediate, structured feedback based on the SEGUE framework was an effective approach for improving anesthesiology students' communication skills. The output generated by large language models like ChatGPT is largely determined by the specificity and quality of the input prompt [ 16 ]. In this study, our prompts, adapted from Wang et al. , constructed a highly realistic and context-specific history-taking process while avoiding ChatGPT's output of professional jargon, premature and excessive answers, and incorrect information [ 9 ]. By instructing ChatGPT to simulate patients with histories of common medical and surgical conditions (e.g., mild hypertension managed with medication) and specific psychosocial concerns (e.g., underlying anxiety about surgery and potential pain), we were able to move beyond generic conversations and achieve professional preoperative visit scenarios. Under the SEGUE framework, trainees in our study were required not only to collect medical information but also to actively protect patient privacy and employ communication skills to address patients' emotions and concerns [ 11 ]. In this study, SEGUE scores were immediately and automatically generated by ChatGPT, producing structured feedback. Unlike delayed summative evaluations from instructors, this integrated feedback loop allowed for true deliberate practice. After each training session, trainees could promptly identify their strengths and areas for improvement. For example, if a trainee failed to inquire about a patient's anxiety, the feedback would directly highlight this gap, enabling immediate cognitive reinforcement and allowing for corrective action in the next practice session. Therefore, the combination teaching method, using a ChatGPT-simulated patient and the SEGUE framework, transforms AI from a mere conversational partner into a convenient educational tool for teaching communication skills. There were several limitations to this study. First, it is impossible to blind participants to their group allocation. Participants in the ChatGPT group were aware they were using a novel AI tool, which may have heightened their motivation or engagement (a Hawthorne effect), potentially contributing to the observed improvement [ 17 ]. We observed that the SEGUE scores of the ChatGPT group increased more rapidly than those of the control group, a pattern that differs from a pure Hawthorne effect, which would be characterized by a higher baseline with parallel improvement trends [ 18 ]. Therefore, the Hawthorne effect likely had a limited impact on our study. Second, a key methodological consideration is the reliability of the SEGUE measure. Although the SEGUE framework is a validated tool, the reliability and consistency of ChatGPT in applying this scoring system across diverse conversations have not been established. It is possible that undiscovered biases or inconsistencies in the AI's scoring algorithm could have influenced our results. However, our ChatGPT-assisted automated scoring method is supported by several studies demonstrating the capability of large language models (LLMs) to reliably automate the scoring of other complex assessments. For example, Quah et al found that the performance of ChatGPT-4 in automated essay scoring for dental undergraduate examinations was strongly correlated with manual scoring, suggesting its potential to aid in self-assessment or large-scale marking automated processes [ 19 ]. Similarly, Jaworski et al r eported that ChatGPT-4.5 demonstrated comparable accuracy to human raters in scoring the Brief International Cognitive Assessment for Multiple Sclerosis (BICAMS) protocols, which include three neuropsychological tests, and was even able to identify scoring errors overlooked by human raters [ 20 ]. This evidence suggests that when provided with a clear, structured scoring standard like SEGUE, ChatGPT is capable of interpreting and applying it to evaluate free-text conversations, supporting its feasibility as a promising tool for the automated assessment of communication skills. Third, although we implemented measures to ensure intervention fidelity, potential limitations related to adherence and contamination should be acknowledged. In this study, Dr. Zhuang was responsible for monitoring students’ training adherence. He met weekly with participants and reviewed their simulated training records. If the participants had not completed the required number of training sessions (at least three times per week), they were asked to complete the remaining sessions immediately. Although these measures were taken to ensure adherence, some variability in the timing and spacing of training sessions across individuals may still have existed, which could have influenced the consistency of the intervention effect. Additionally, as in most educational trials, contamination between participants is a concern [ 21 ]. Unlike internal medicine or surgery departments, where trainees often work in shared spaces, anesthesiology students work independently in physically separate operating rooms. This physical separation likely minimized opportunities for contamination between groups. However, it remains impossible to guarantee that no communication occurred between participants in the two groups, and it would be ethically inappropriate to prohibit students from discussing their learning experiences. Fourth, our study assessed communication skills in an AI-simulated environment. Whether the observed improvements translate directly to enhanced performance and patient outcomes in real-world clinical settings remains a critical area for future research. Furthermore, these findings may not be directly generalizable to other medical centers, different levels of trainees, or other departments. Future multicenter studies with diverse populations are needed to confirm the external validity of our results. Conclusions This study demonstrated that a one-month training program combining a ChatGPT-simulated patient with automated SEGUE feedback significantly improved the preoperative communication skills of anesthesiology students. Despite the need for further validation in real-world clinical settings, our findings support the integration of large language models into medical education to enhance the communication skills of medical students. Abbreviations AI Artificial Intelligence ChatGPT Chat Generative Pre-trained Transformer CI Confidence Interval LLM Large Language Model RCT Randomized Controlled Trial SD Standard Deviation SEGUE Set the stage, Elicit information, Give information, Understand the patient's perspective, End the encounter Declarations Supplementary Information The online version contains supplementary material available at Acknowledgements None. Author contributions P.Z. and J.Z. conceived the study idea; P.Z., Y.L. and J.Z. analyzed the study data. All authors drafted and revised the manuscript. The final manuscript was read and approved by all the authors. Funding None. Data availability The trial protocol and statistical analysis plan are available from Chinese Clinical Trial Registry (https://www.chictr.org.cn/, ID: ChiCTR2500106474). The data generated during this study are not publicly available due to privacy regulations and limitations on the informed consent of participants, but are available from the corresponding author (Dr. Jing Zhuang) on reasonable request. Ethics approval and consent to participate This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Clinical and Biomedical Trials at Sichuan Provincial People's Hospital (No. 2025-389). Informed consent was obtained from all individual participants in the study. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Author details Department of Anesthesiology, Sichuan Provincial People’s Hospital, School of Medicine, University of Electronic Science and Technology of China, 32 West Second Section First Ring Road, Chengdu, Sichuan, China References Yedidia MJ, Gillespie CC, Kachur E, Schwartz MD, Ockene J, Chepaitis AE, Snyder CW, Lazare A, Lipkin M. Jr. Effect of communications training on medical student performance. JAMA. 2003;290(9):1157–65. Lane C, Rollnick S. The use of simulated patients and role-play in communication skills training: a review of the literature to August 2005. Patient Educ Couns. 2007;67(1–2):13–20. Gelis A, Cervello S, Rey R, Llorca G, Lambert P, Franck N, Dupeyron A, Delpont M, Rolland B. Peer Role-Play for Training Communication Skills in Medical Students: A Systematic Review. Simul Healthc. 2020;15(2):106–11. Nikendei C, Kraus B, Schrauth M, Weyrich P, Zipfel S, Herzog W, Jünger J. Integration of role-playing into technical skills training: a randomized controlled trial. Med Teach. 2007;29(9):956–60. Anderson MB, Stillman PL, Wang Y. Growing use of standardized patients in teaching and evaluation in medical education. Teach Learn Med. 1994;6(1):15–22. Barrows HS. An overview of the uses of standardized patients for teaching and evaluating clinical skills. AAMC. Acad Med. 1993;68(6):443–51. discussion 51 – 3. Bosse HM, Nickel M, Huwendiek S, Schultz JH, Nikendei C. Cost-effectiveness of peer role play and standardized patients in undergraduate communication training. BMC Med Educ. 2015;15:183. Webb JJ. Proof of Concept: Using ChatGPT to Teach Emergency Physicians How to Break Bad News. Cureus. 2023;15(5):e38755. Wang C, Li S, Lin N, Zhang X, Han Y, Wang X, Liu D, Tan X, Pu D, Li K, et al. Application of Large Language Models in Medical Training Evaluation-Using ChatGPT as a Standardized Patient: Multimetric Assessment. J Med Internet Res. 2025;27:e59435. Hopewell S, Chan AW, Collins GS, Hróbjartsson A, Moher D, Schulz KF, Tunn R, Aggarwal R, Berkwits M, Berlin JA, et al. CONSORT 2025 statement: updated guideline for reporting randomised trials. BMJ. 2025;389:e081123. Makoul G. The SEGUE Framework for teaching and assessing communication skills. Patient Educ Couns. 2001;45(1):23–34. Zeng N, Lu H, Li S, Yang Q, Liu F, Pan H, Yan S. Application of the combination of CBL teaching method and SEGUE framework to improve the doctor-patient communication skills of resident physicians in otolaryngology department. BMC Med Educ. 2024;24(1):201. Rederiene G, Buunk-Werkhoven Y, Aidukaite G, Puriene A. Relationship Between Job Satisfaction and Health of Hygienists in Lithuania. Int Dent J. 2022;72(4):512–18. García-Torres D, Vicente Ripoll MA, Fernández Peris C, Mira Solves JJ. Enhancing Clinical Reasoning with Virtual Patients: A Hybrid Systematic Review Combining Human Reviewers and ChatGPT. Healthc (Basel). 2024;12(22). Klafta JM, Roizen MF. Current understanding of patients' attitudes toward and preparation for anesthesia: a review. Anesth Analg. 1996;83(6):1314–21. Meskó B. Prompt Engineering as an Important Emerging Skill for Medical Professionals: Tutorial. J Med Internet Res. 2023;25:e50638. Demetriou C, Hu L, Smith TO, Hing CB. Hawthorne effect on surgical studies. ANZ J Surg. 2019;89(12):1567–76. Adair JG. The Hawthorne effect: A reconsideration of the methodological artifact. J Appl Psychol. 1984;69(2):334–45. Quah B, Zheng L, Sng TJH, Yong CW, Islam I. Reliability of ChatGPT in automated essay scoring for dental undergraduate examinations. BMC Med Educ. 2024;24(1):962. Jaworski M 3rd, Balconi J, Santivasci C, Calamia M. Feasibility of AI-powered assessment scoring: Can large language models replace human raters? Clin Neuropsychol. 2025:1–14. Keogh-Brown MR, Bachmann MO, Shepstone L, Hewitt C, Howe A, Ramsay CR, Song F, Miles JN, Torgerson DJ, Miles S et al. Contamination in trials of educational interventions. Health Technol Assess. 2007;11(43):iii, ix-107. Additional Declarations No competing interests reported. Supplementary Files 2supplementaryfile2CONSORT2025editablechecklist.docx 2supplementaryfile2.xlsx Cite Share Download PDF Status: Under Review Version 1 posted Reviewers invited by journal 25 Mar, 2026 Editor assigned by journal 23 Mar, 2026 Editor invited by journal 03 Mar, 2026 Submission checks completed at journal 26 Feb, 2026 First submitted to journal 26 Feb, 2026 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-8898929","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":612238462,"identity":"0c9e666e-13f0-4dae-b86d-232795a8d1c1","order_by":0,"name":"Peng Zhang","email":"","orcid":"","institution":"Sichuan Academy of Medical Sciences \u0026 Sichuan Provincial People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Peng","middleName":"","lastName":"Zhang","suffix":""},{"id":612238463,"identity":"cffd18be-4db5-49c1-8a0a-c3bf525c1706","order_by":1,"name":"Yu Liu","email":"","orcid":"","institution":"Sichuan Academy of Medical Sciences \u0026 Sichuan Provincial People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yu","middleName":"","lastName":"Liu","suffix":""},{"id":612238464,"identity":"acc95c82-ef62-4160-b442-0f0e6e6ed208","order_by":2,"name":"Jing Zhuang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4UlEQVRIiWNgGAWjYBACAwhlw8PG3nzwQUJFDdFa0mT4eI4lGzw4c4xoLYdt5CRyzCQftjAT1mLO3nvsMw/DYR42hrS0isQGNgb+9u4EvFose84lz+ZhSAdqOXzsRuIOGQaJM2c34HfYjRxjZh4Gax42xra0G4ln2BgMJHIJaLn/BqSFmYeNmcesILGNmQgtN3hAWpx52Nh4zBiI03Imx5hxDkMaDxsPW7JEwpljPIT9cvyMMcMbBht7+fmPD378UVEjx9/ei18LCDDx/kNweAgqBwHGH0QpGwWjYBSMghELAPA6QG6LRZ/cAAAAAElFTkSuQmCC","orcid":"","institution":"Sichuan Academy of Medical Sciences \u0026 Sichuan Provincial People's Hospital","correspondingAuthor":true,"prefix":"","firstName":"Jing","middleName":"","lastName":"Zhuang","suffix":""}],"badges":[],"createdAt":"2026-02-17 08:23:05","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8898929/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8898929/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105571293,"identity":"92dd5d3b-b484-4a7c-91ea-5b19bd75217f","added_by":"auto","created_at":"2026-03-27 13:22:25","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":352047,"visible":true,"origin":"","legend":"\u003cp\u003eDiagram of patient flow\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8898929/v1/89ca78df66a424ebe35fde73.jpeg"},{"id":105573494,"identity":"f1b981f9-7c91-4e14-bd2c-b1fa53395ab9","added_by":"auto","created_at":"2026-03-27 13:31:36","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":982418,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8898929/v1/685a3e85-7751-4992-8368-84b2bb8db26a.pdf"},{"id":105571550,"identity":"3dab476a-13fc-4cc4-b3cf-9061c1a7560e","added_by":"auto","created_at":"2026-03-27 13:23:30","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":36844,"visible":true,"origin":"","legend":"","description":"","filename":"2supplementaryfile2CONSORT2025editablechecklist.docx","url":"https://assets-eu.researchsquare.com/files/rs-8898929/v1/cb2d4d927db8d91830b0b3f7.docx"},{"id":105571578,"identity":"8e32f163-c5ae-40ff-bdb2-f568c5c57953","added_by":"auto","created_at":"2026-03-27 13:23:40","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":10372,"visible":true,"origin":"","legend":"","description":"","filename":"2supplementaryfile2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-8898929/v1/38acdb80199f189144b10cb2.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Teaching communication skills to anesthesiology students using ChatGPT as a simulated patient and the SEGUE framework","fulltext":[{"header":"Introduction","content":"\u003cp\u003eAnesthesiologists, despite often being perceived as physicians who operate behind the scenes, still require robust patient communication skills, particularly during the preoperative visit process. Preoperative anesthesiologist-patient interactions are frequently complicated by several factors, including time constraints, patient anxiety, a noisy bedside environment, and absence of a pre-existing relationship. Traditional medical education primarily focuses on teaching students basic medical theory, but tends to neglect the development of doctor-patient communication skills [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. This situation often results in medical students not feeling confident enough to communicate with patients in real clinical settings.\u003c/p\u003e \u003cp\u003eStandardized patients and peer role-playing are two fundamental simulation-based medical education methods for training technical (procedural) skills and communication skills in medical students [\u003cspan additionalcitationids=\"CR3\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Simulated patients, who are professional actors or trained individuals, are used to teach communication skills by delivering standardized and customizable scenarios and providing feedback [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. While they offer a controlled learning environment, their use can be expensive, time-consuming, and occasionally inconsistent despite training [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Role-play among medical students is a cost-effective way to practice communication skills in a simulated setting [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. However, it may feel artificial and encounter resistance from trainees who are reluctant to take part in role-play with colleagues [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eChatGPT, developed by OpenAI, is capable of interacting with individual users and generating human-like responses to inputs in a conversational style. This technology has been documented in the medical education. Moreover, some authors have reported its novel use as a role-playing device to generate human-like responses in fictional but realistic clinical scenarios for educational purposes. \u003cem\u003eWebb et al.\u003c/em\u003e have utilized ChatGPT-3.5 as an educational tool to train emergency physicians in the breaking bad news to patients [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. \u003cem\u003eWang et al.\u003c/em\u003e have demonstrated that ChatGPT can effectively simulate a standardized patient by inputting appropriate prompts, showing its potential to improve medical training [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn this study, we aimed to use ChatGPT-4.0 to design a training scenario in which anesthesiology students would engage in role-play tasks with a simulated patient to practice preoperative assessments. Additionally, we also aimed to evaluate whether the communication skills of the students improved following the one month training.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eEthics consideration\u003c/h2\u003e \u003cp\u003e This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Clinical and Biomedical Trials at Sichuan Provincial People's Hospital (Approval No. 2025\u0026thinsp;\u0026minus;\u0026thinsp;389). Each participant provided written informed consent. This trial was registered prior to patient enrollment with the Chinese Clinical Trial Registry (Registration Number: ChiCTR2500106474). The principal investigator for the registration was Dr. Jing Zhuang, and the date of registration was 2025-07-24. No patient or public was involved in the design, conduct, or reporting of this study. This randomized controlled trial is reported in accordance with the CONSORT (Consolidated Standards of Reporting Trials) 2025 guidelines (supplementary file 1) [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eStudy participants\u003c/h3\u003e\n\u003cp\u003eThis parallel-group randomized trial was conducted at Sichuan Provincial People's Hospital, Chengdu, Sichuan, China. No changes were made to the trial protocol after commencement. The inclusion criteria were as follows: anesthesiology students (including postgraduate students and residents) who are scheduled for a clinical rotation in the Department of Anesthesiology, Sichuan Provincial People's Hospital, from July 2025 to March 2026. The exclusion criteria were as follows: (1) individuals unable to operate electronic devices (e.g., computers, tablets, or smartphones); and (2) individuals without internet access. The allocation sequence was generated using a computer software (SPSS, version 26.0) by an independent researcher using simple randomization, with no stratification or blocking. An independent researcher sealed the sequence in consecutively numbered, opaque, sealed envelopes to conceal the allocation. The investigators responsible for participant enrollment and intervention were kept unaware of the sequence. This study recruited a cohort of 34 participants. All participants have completed both fundamental medical theory courses and clinical practice components at the undergraduate level. Participants were not blinded to their group allocation due to the visible difference in receiving SEGUE feedback.\u003c/p\u003e\n\u003ch3\u003eChatGPT-4 simulated as a standardized patent\u003c/h3\u003e\n\u003cp\u003eThe initial input prompt was designed to establish a medical scenario involving anesthesiology, with ChatGPT assigned the role of the patient. This prompt was adapted from a study by \u003cem\u003eWang et al.\u003c/em\u003e [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e], which used ChatGPT as a standardized patient for history-taking tasks. The effectiveness of this prompt have been validated based on its anthropomorphism, clinical accuracy, and adaptability [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Acting as the anesthesiologist, the trainees provided additional inputs in response to ChatGPT's outputs to simulate a realistic and natural conversation during the preoperative visit process. The input and output responses were recorded using screen capture software. The input prompts used to design ChatGPT-4 as the simulated patient are detailed in supplementary file 2.\u003c/p\u003e\n\u003ch3\u003eSEGUE framework\u003c/h3\u003e\n\u003cp\u003eThe SEGUE framework, developed by Professor \u003cem\u003eMakoul\u003c/em\u003e at the Feinberg School of Medicine, Northwestern University, is designed to guide medical students in communicating effectively with patients [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. SEGUE is an acronym for its five core components: set the stage, elicit information, give information, understand the patient\u0026rsquo;s perspective and end the encounter [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. The SEGUE framework employs a binary scoring system, where each item was rated as either \"Yes\" (score\u0026thinsp;=\u0026thinsp;1) or \"No\" (score\u0026thinsp;=\u0026thinsp;0) [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. The total score for each encounter was calculated by summing the scores for each item, with higher scores indicating better provider-patient communication skills [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. We used a Chinese version of the SEGUE to measure the communication performance of the trainees [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e].\u003c/p\u003e\n\u003ch3\u003eOverall satisfaction evaluation\u003c/h3\u003e\n\u003cp\u003eOverall satisfaction was measured using a 5-point Likert scale: 1\u0026thinsp;=\u0026thinsp;completely dissatisfied, 2\u0026thinsp;=\u0026thinsp;dissatisfied, 3\u0026thinsp;=\u0026thinsp;partially satisfied, 4\u0026thinsp;=\u0026thinsp;satisfied, and 5\u0026thinsp;=\u0026thinsp;completely satisfied. Higher total scores indicate greater respondent satisfaction [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e].\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eStudy procedure\u003c/h2\u003e \u003cp\u003e Participants in the ChatGPT-assisted group underwent a structured training protocol, beginning with an initial ChatGPT-simulated preoperative assessment, after which ChatGPT-4 automatically generated a SEGUE score and provided structured feedback based on the SEGUE framework. Over the subsequent month, they were required to independently complete this simulated assessment at least three times per week, each followed by real-time SEGUE feedback. In contrast, control group participants were assessed using the same ChatGPT system at baseline and the study endpoint but received no SEGUE feedback or additional training. Participants were not blinded to their group allocation. All SEGUE scoring was performed automatically by ChatGPT-4 using de-identified conversation transcripts. Endpoint assessments employed parallel case scenarios. Finally, an overall satisfaction evaluation (using a 5-point Likert scale) was administered by a research assistant who was blinded to the group assignments.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eOutcomes\u003c/h3\u003e\n\u003cp\u003eThe primary outcome was the change in SEGUE scores from baseline to one-month follow-up, measured as the mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation for each group. The secondary outcome was overall satisfaction with the ChatGPT-assisted training method, assessed using a 5-point Likert scale at the end of the study period. No harms or adverse events were assessed, as the intervention was educational. Outcome assessment was performed by a research assistant who was blinded to group assignments. The data analyst was also blinded to group allocation.\u003c/p\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eStatistical analyses were performed using SPSS (Version 26.0; IBM Corp., Armonk, NY, USA). The sample size was determined using G*Power (Version 3.1.9.7; Heinrich-Heine-Universit\u0026auml;t D\u0026uuml;sseldorf, D\u0026uuml;sseldorf, Germany). Based on an anticipated medium effect size (f\u0026thinsp;=\u0026thinsp;0.25), an α level of 0.05, and a desired power of 0.8, the calculated total sample size was 34 participants. No interim analyses were planned or conducted. Continuous variables are presented as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (SD), and categorical data are reported as numbers (%). Baseline differences between the two groups in terms of sex, age, and education level were assessed using independent samples t-tests. A repeated-measures analysis of variance (ANOVA) was employed to compare the SEGUE scores between groups across two time points (baseline and one-month follow-up). All analyses were performed on an intention-to-treat basis, including all randomized participants. A p-value of less than 0.05 was considered statistically significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003eParticipants were recruited from August 2025 to January 2026, with the one-month follow-up completed by February 2026. A total of 98 anesthesiology students scheduled for their clinical rotation in the Department of Anesthesiology were assessed for eligibility and met the inclusion criteria. Of these, 64 were excluded: 58 because they were not assigned to Dr. Zhang for clinical supervision during the study period, 2 because they lacked electronic devices or internet access, and 4 because they declined to participate. Finally, 34 students were enrolled and randomized to the ChatGPT group (n\u0026thinsp;=\u0026thinsp;17) or the control group (n\u0026thinsp;=\u0026thinsp;17). All 34 participants received the intended intervention and were included in the primary outcome analysis. There were no losses to follow-up or exclusions after randomization. The participant flow diagram is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eThe ChatGPT group comprised five male and 12 female participants (mean age\u0026thinsp;=\u0026thinsp;24.06\u0026thinsp;\u0026plusmn;\u0026thinsp;0.83 years), while the control group consisted of eight male and nine female participants (mean age\u0026thinsp;=\u0026thinsp;24.41\u0026thinsp;\u0026plusmn;\u0026thinsp;0.71 years). There was no significant difference in sex (p\u0026thinsp;=\u0026thinsp;0.481) and age (p\u0026thinsp;=\u0026thinsp;0.192) between the two groups.\u003c/p\u003e \u003cp\u003eParticipants in the ChatGPT group completed a mean of 13.18\u0026thinsp;\u0026plusmn;\u0026thinsp;1.38 (range 12 to 16) simulated preoperative assessment training during the one-month period. No participants in the control group reported engaging in any additional communication skills training during the study period.\u003c/p\u003e \u003cp\u003eMain effect of time: The within-subjects effect test indicated a significant time effect on SEGUE scores (F\u0026thinsp;=\u0026thinsp;95.74, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), demonstrating statistically significant differences in SEGUE scores across different time points. Time \u0026times; Group interaction: The interaction between time and group was significant (F\u0026thinsp;=\u0026thinsp;44.51, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), suggesting different trends of SEGUE score changes over time between groups. Specifically, the ChatGPT group showed significantly faster score improvement than the control group. Between-group main effect: The between-subjects effect test revealed statistically significant differences in overall SEGUE scores between groups (F\u0026thinsp;=\u0026thinsp;13.93, p\u0026thinsp;=\u0026thinsp;0.01), indicating that the intervention effects on SEGUE scores differed between groups, with the ChatGPT group demonstrating superior intervention effectiveness compared to the control group (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe baseline SEGUE scores were comparable between the two groups (p\u0026thinsp;=\u0026thinsp;0.904). Both groups exhibited significant improvements in SEGUE scores compared to their baseline values (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001 for the ChatGPT group and p\u0026thinsp;=\u0026thinsp;0.035 for the control group). Notably, the SEGUE scores of the ChatGPT group were significantly higher than those of the control group after three months of training (22.29\u0026thinsp;\u0026plusmn;\u0026thinsp;1.36 vs. 16.88\u0026thinsp;\u0026plusmn;\u0026thinsp;2.47, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOverall satisfaction was assessed only in the ChatGPT group. Analysis of the survey data revealed favorable acceptance of the ChatGPT-assisted teaching methodology in the ChatGPT group: no participants (0%) expressed complete dissatisfaction, one participant (6%) expressed dissatisfaction, three participants (17%) expressed partial satisfaction, eight participants (47%) expressed satisfaction, and five participants (30%) expressed complete satisfaction (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eResults of repeated measures ANOVA of SEGUE scores\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eSEGUE scores before training\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eChatGPT group\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eControl group\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eF\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e15.76\u0026thinsp;\u0026plusmn;\u0026thinsp;2.86\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e15.65\u0026thinsp;\u0026plusmn;\u0026thinsp;2.81\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.15\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.904\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSEGUE scores after training\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e22.29\u0026thinsp;\u0026plusmn;\u0026thinsp;1.36\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e16.88\u0026thinsp;\u0026plusmn;\u0026thinsp;2.47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e62.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e135.40\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.035\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBetween-group main effect\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eF\u0026thinsp;=\u0026thinsp;13.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eP\u0026thinsp;=\u0026thinsp;0.01\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMain effect of time\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eF\u0026thinsp;=\u0026thinsp;95.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eP\u0026thinsp;\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTime \u0026times; Group interaction\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eF\u0026thinsp;=\u0026thinsp;44.51\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eP\u0026thinsp;\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eResidents\u0026rsquo; satisfaction evaluation of ChatGPT training\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSatisfaction evaluation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNumber of respondents\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePercentage\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCompletely dissatisfied\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026mdash;\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDissatisfied\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePartially satisfied\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e17%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSatisfied\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e47%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCompletely satisfied\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e30%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we investigated a convenient and interactive method to teach communication skills to anesthesiology students. We found that a one-month training program with a combination teaching method, using a ChatGPT-simulated patient and the SEGUE framework, significantly improved the preoperative communication skills of anesthesiology trainees. Furthermore, the overall satisfaction survey indicated that the majority of participants found this AI-assisted teaching method acceptable and beneficial. These findings highlight the potential of large language models like ChatGPT as a flexible and effective tool for simulating clinical interactions and enhancing communication skills training in medical education.\u003c/p\u003e \u003cp\u003eOur findings reinforce a small but growing body of literature exploring the use of ChatGPT in clinical role-playing [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. For instance, \u003cem\u003eWebb et al.\u003c/em\u003e and \u003cem\u003eWang et al.\u003c/em\u003e have reported its successful application in simulating patients for training in breaking bad news and in history-taking tasks, respectively [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. The present study extends these findings by applying this technology specifically to the anesthesiology preoperative assessment\u0026mdash;a context characterized by time pressure and the need to rapidly build trust and gather complex information [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Compared to the control group, our results showed that the ChatGPT group's SEGUE scores increased more rapidly (as reflected in the significant Time \u0026times; Group interaction effect) and reached higher levels. This outcome demonstrates that the iterative practice (three times per week) with immediate, structured feedback based on the SEGUE framework was an effective approach for improving anesthesiology students' communication skills.\u003c/p\u003e \u003cp\u003eThe output generated by large language models like ChatGPT is largely determined by the specificity and quality of the input prompt [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. In this study, our prompts, adapted from \u003cem\u003eWang et al.\u003c/em\u003e, constructed a highly realistic and context-specific history-taking process while avoiding ChatGPT's output of professional jargon, premature and excessive answers, and incorrect information [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. By instructing ChatGPT to simulate patients with histories of common medical and surgical conditions (e.g., mild hypertension managed with medication) and specific psychosocial concerns (e.g., underlying anxiety about surgery and potential pain), we were able to move beyond generic conversations and achieve professional preoperative visit scenarios.\u003c/p\u003e \u003cp\u003eUnder the SEGUE framework, trainees in our study were required not only to collect medical information but also to actively protect patient privacy and employ communication skills to address patients' emotions and concerns [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. In this study, SEGUE scores were immediately and automatically generated by ChatGPT, producing structured feedback. Unlike delayed summative evaluations from instructors, this integrated feedback loop allowed for true deliberate practice. After each training session, trainees could promptly identify their strengths and areas for improvement. For example, if a trainee failed to inquire about a patient's anxiety, the feedback would directly highlight this gap, enabling immediate cognitive reinforcement and allowing for corrective action in the next practice session. Therefore, the combination teaching method, using a ChatGPT-simulated patient and the SEGUE framework, transforms AI from a mere conversational partner into a convenient educational tool for teaching communication skills.\u003c/p\u003e \u003cp\u003eThere were several limitations to this study. First, it is impossible to blind participants to their group allocation. Participants in the ChatGPT group were aware they were using a novel AI tool, which may have heightened their motivation or engagement (a Hawthorne effect), potentially contributing to the observed improvement [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. We observed that the SEGUE scores of the ChatGPT group increased more rapidly than those of the control group, a pattern that differs from a pure Hawthorne effect, which would be characterized by a higher baseline with parallel improvement trends [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Therefore, the Hawthorne effect likely had a limited impact on our study. Second, a key methodological consideration is the reliability of the SEGUE measure. Although the SEGUE framework is a validated tool, the reliability and consistency of ChatGPT in applying this scoring system across diverse conversations have not been established. It is possible that undiscovered biases or inconsistencies in the AI's scoring algorithm could have influenced our results. However, our ChatGPT-assisted automated scoring method is supported by several studies demonstrating the capability of large language models (LLMs) to reliably automate the scoring of other complex assessments. For example, \u003cem\u003eQuah et al\u003c/em\u003e found that the performance of ChatGPT-4 in automated essay scoring for dental undergraduate examinations was strongly correlated with manual scoring, suggesting its potential to aid in self-assessment or large-scale marking automated processes [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Similarly, \u003cem\u003eJaworski et al r\u003c/em\u003eeported that ChatGPT-4.5 demonstrated comparable accuracy to human raters in scoring the Brief International Cognitive Assessment for Multiple Sclerosis (BICAMS) protocols, which include three neuropsychological tests, and was even able to identify scoring errors overlooked by human raters [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. This evidence suggests that when provided with a clear, structured scoring standard like SEGUE, ChatGPT is capable of interpreting and applying it to evaluate free-text conversations, supporting its feasibility as a promising tool for the automated assessment of communication skills. Third, although we implemented measures to ensure intervention fidelity, potential limitations related to adherence and contamination should be acknowledged. In this study, Dr. Zhuang was responsible for monitoring students\u0026rsquo; training adherence. He met weekly with participants and reviewed their simulated training records. If the participants had not completed the required number of training sessions (at least three times per week), they were asked to complete the remaining sessions immediately. Although these measures were taken to ensure adherence, some variability in the timing and spacing of training sessions across individuals may still have existed, which could have influenced the consistency of the intervention effect. Additionally, as in most educational trials, contamination between participants is a concern [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Unlike internal medicine or surgery departments, where trainees often work in shared spaces, anesthesiology students work independently in physically separate operating rooms. This physical separation likely minimized opportunities for contamination between groups. However, it remains impossible to guarantee that no communication occurred between participants in the two groups, and it would be ethically inappropriate to prohibit students from discussing their learning experiences. Fourth, our study assessed communication skills in an AI-simulated environment. Whether the observed improvements translate directly to enhanced performance and patient outcomes in real-world clinical settings remains a critical area for future research. Furthermore, these findings may not be directly generalizable to other medical centers, different levels of trainees, or other departments. Future multicenter studies with diverse populations are needed to confirm the external validity of our results.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThis study demonstrated that a one-month training program combining a ChatGPT-simulated patient with automated SEGUE feedback significantly improved the preoperative communication skills of anesthesiology students. Despite the need for further validation in real-world clinical settings, our findings support the integration of large language models into medical education to enhance the communication skills of medical students.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eAI\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Artificial Intelligence\u003c/p\u003e\n\u003cp\u003eChatGPT\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Chat Generative Pre-trained Transformer\u003c/p\u003e\n\u003cp\u003eCI\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Confidence Interval\u003c/p\u003e\n\u003cp\u003eLLM\u0026nbsp; \u0026nbsp;\u0026nbsp;\u0026nbsp; \u0026nbsp;Large Language Model\u003c/p\u003e\n\u003cp\u003eRCT\u0026nbsp; \u0026nbsp;\u0026nbsp;\u0026nbsp; \u0026nbsp;Randomized Controlled Trial\u003c/p\u003e\n\u003cp\u003eSD\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Standard Deviation\u003c/p\u003e\n\u003cp\u003eSEGUE \u0026nbsp; \u0026nbsp; \u0026nbsp; Set the stage, Elicit information, Give information, Understand the \u0026nbsp; \u0026nbsp; \u0026nbsp;patient's perspective, End the encounter\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eSupplementary Information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe online version contains supplementary material available at\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNone.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eP.Z. and J.Z. conceived the study idea; P.Z., Y.L. and J.Z. analyzed the study data. All authors drafted and revised the manuscript. The final manuscript was read and approved by all the authors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNone.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe trial protocol and statistical analysis plan are available from Chinese Clinical Trial Registry (https://www.chictr.org.cn/, ID: ChiCTR2500106474). The data generated during this study are not publicly available due to privacy regulations and limitations on the informed consent of participants, but are available from the corresponding author (Dr. Jing Zhuang) on reasonable request.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Clinical and Biomedical Trials at Sichuan Provincial People's Hospital (No. 2025-389). Informed consent was obtained from all individual participants in the study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor details\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDepartment of Anesthesiology, Sichuan Provincial People’s Hospital, School of Medicine, University of Electronic Science and Technology of China, 32 West Second Section First Ring Road, Chengdu, Sichuan, China\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eYedidia MJ, Gillespie CC, Kachur E, Schwartz MD, Ockene J, Chepaitis AE, Snyder CW, Lazare A, Lipkin M. Jr. Effect of communications training on medical student performance. JAMA. 2003;290(9):1157\u0026ndash;65.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLane C, Rollnick S. The use of simulated patients and role-play in communication skills training: a review of the literature to August 2005. Patient Educ Couns. 2007;67(1\u0026ndash;2):13\u0026ndash;20.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGelis A, Cervello S, Rey R, Llorca G, Lambert P, Franck N, Dupeyron A, Delpont M, Rolland B. Peer Role-Play for Training Communication Skills in Medical Students: A Systematic Review. Simul Healthc. 2020;15(2):106\u0026ndash;11.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNikendei C, Kraus B, Schrauth M, Weyrich P, Zipfel S, Herzog W, J\u0026uuml;nger J. Integration of role-playing into technical skills training: a randomized controlled trial. Med Teach. 2007;29(9):956\u0026ndash;60.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnderson MB, Stillman PL, Wang Y. Growing use of standardized patients in teaching and evaluation in medical education. Teach Learn Med. 1994;6(1):15\u0026ndash;22.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarrows HS. An overview of the uses of standardized patients for teaching and evaluating clinical skills. AAMC. Acad Med. 1993;68(6):443\u0026ndash;51. discussion 51\u0026thinsp;\u0026ndash;\u0026thinsp;3.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBosse HM, Nickel M, Huwendiek S, Schultz JH, Nikendei C. Cost-effectiveness of peer role play and standardized patients in undergraduate communication training. BMC Med Educ. 2015;15:183.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWebb JJ. Proof of Concept: Using ChatGPT to Teach Emergency Physicians How to Break Bad News. Cureus. 2023;15(5):e38755.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang C, Li S, Lin N, Zhang X, Han Y, Wang X, Liu D, Tan X, Pu D, Li K, et al. Application of Large Language Models in Medical Training Evaluation-Using ChatGPT as a Standardized Patient: Multimetric Assessment. J Med Internet Res. 2025;27:e59435.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHopewell S, Chan AW, Collins GS, Hr\u0026oacute;bjartsson A, Moher D, Schulz KF, Tunn R, Aggarwal R, Berkwits M, Berlin JA, et al. CONSORT 2025 statement: updated guideline for reporting randomised trials. BMJ. 2025;389:e081123.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMakoul G. The SEGUE Framework for teaching and assessing communication skills. Patient Educ Couns. 2001;45(1):23\u0026ndash;34.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZeng N, Lu H, Li S, Yang Q, Liu F, Pan H, Yan S. Application of the combination of CBL teaching method and SEGUE framework to improve the doctor-patient communication skills of resident physicians in otolaryngology department. BMC Med Educ. 2024;24(1):201.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRederiene G, Buunk-Werkhoven Y, Aidukaite G, Puriene A. Relationship Between Job Satisfaction and Health of Hygienists in Lithuania. Int Dent J. 2022;72(4):512\u0026ndash;18.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarc\u0026iacute;a-Torres D, Vicente Ripoll MA, Fern\u0026aacute;ndez Peris C, Mira Solves JJ. Enhancing Clinical Reasoning with Virtual Patients: A Hybrid Systematic Review Combining Human Reviewers and ChatGPT. Healthc (Basel). 2024;12(22).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKlafta JM, Roizen MF. Current understanding of patients' attitudes toward and preparation for anesthesia: a review. Anesth Analg. 1996;83(6):1314\u0026ndash;21.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMesk\u0026oacute; B. Prompt Engineering as an Important Emerging Skill for Medical Professionals: Tutorial. J Med Internet Res. 2023;25:e50638.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDemetriou C, Hu L, Smith TO, Hing CB. Hawthorne effect on surgical studies. ANZ J Surg. 2019;89(12):1567\u0026ndash;76.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAdair JG. The Hawthorne effect: A reconsideration of the methodological artifact. J Appl Psychol. 1984;69(2):334\u0026ndash;45.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQuah B, Zheng L, Sng TJH, Yong CW, Islam I. Reliability of ChatGPT in automated essay scoring for dental undergraduate examinations. BMC Med Educ. 2024;24(1):962.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJaworski M 3rd, Balconi J, Santivasci C, Calamia M. Feasibility of AI-powered assessment scoring: Can large language models replace human raters? Clin Neuropsychol. 2025:1\u0026ndash;14.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeogh-Brown MR, Bachmann MO, Shepstone L, Hewitt C, Howe A, Ramsay CR, Song F, Miles JN, Torgerson DJ, Miles S et al. Contamination in trials of educational interventions. Health Technol Assess. 2007;11(43):iii, ix-107.\u003c/span\u003e\u003c/li\u003e\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-anesthesiology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bane","sideBox":"Learn more about [BMC Anesthesiology](http://bmcanesthesiol.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bane","title":"BMC Anesthesiology","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"ChatGPT, medical education, AI, medical research","lastPublishedDoi":"10.21203/rs.3.rs-8898929/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8898929/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eIn this randomized controlled trial, we aimed to evaluate the efficacy of a one-month training program using ChatGPT-simulated patient and automated SEGUE framework feedback in improving the preoperative communication skills of anesthesiology students.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThirty-four anesthesiology students were randomly allocated to either a ChatGPT group (n\u0026thinsp;=\u0026thinsp;17) or a control group (n\u0026thinsp;=\u0026thinsp;17). The ChatGPT group underwent a structured training program, conducting at least three simulated preoperative assessments per week with ChatGPT-4.0, which provided immediate SEGUE scores and feedback. The control group received only baseline and endpoint assessments with ChatGPT-4.0 as well, but without SEGUE feedback. The primary outcome was the change in SEGUE scores, which range from 0 to 25.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe ChatGPT group showed a faster improvement trend in SEGUE scores than the control group, suggesting a significant Time \u0026times; Group interaction effect (F\u0026thinsp;=\u0026thinsp;44.51, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The ChatGPT group obtained higher SEGUE scores (from 15.76\u0026thinsp;\u0026plusmn;\u0026thinsp;2.86 to 22.29\u0026thinsp;\u0026plusmn;\u0026thinsp;1.36) compared to the control group (from 15.65\u0026thinsp;\u0026plusmn;\u0026thinsp;2.81 to 16.88\u0026thinsp;\u0026plusmn;\u0026thinsp;2.47) after one month (p\u0026thinsp;=\u0026thinsp;0.01). Satisfaction survey indicated that 94% of the ChatGPT group participants found the method acceptable (47% satisfied, 30% completely satisfied).\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eA ChatGPT-assisted training program significantly enhanced the preoperative communication skills of anesthesiology students and was well-received. This approach represents a promising, scalable tool for complementing traditional communication skills training in medical education.\u003c/p\u003e\u003ch2\u003eTrial registration:\u003c/h2\u003e \u003cp\u003eThis study was registered in the Chinese Clinical Trial Registry (ChiCTR2500106474) on July 24, 2025.\u003c/p\u003e","manuscriptTitle":"Teaching communication skills to anesthesiology students using ChatGPT as a simulated patient and the SEGUE framework","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-27 12:25:21","doi":"10.21203/rs.3.rs-8898929/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewersInvited","content":"","date":"2026-03-25T05:46:09+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-23T07:53:05+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-03-03T05:59:45+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-02-26T17:40:36+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Anesthesiology","date":"2026-02-26T11:20:00+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-anesthesiology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bane","sideBox":"Learn more about [BMC Anesthesiology](http://bmcanesthesiol.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bane","title":"BMC Anesthesiology","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"87bb8030-c9bd-4e3b-8dcb-e8522acc1351","owner":[],"postedDate":"March 27th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-03-27T12:25:21+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-27 12:25:21","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8898929","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8898929","identity":"rs-8898929","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}