Real-World Analysis of Organ Transplantation-Specific Agent Based on Large Language Model in Post-Transplant Self-Management During Off-Hours: A Mixed-Methods Study

Abstract

Objective

A significant gap exists in medical support for organ transplant patients during out-of-hours (OOH). General large language models (LLMs), affected by AI hallucinations, are unsuitable for complex post-transplant care. We built the first post-transplant AI agent based on LLMs to address these issues.

Methods

We constructed a specialized “post-transplant AI agent” (named Doctor Xiao Yi) and conducted a mixed-methods study comparing it to a hospital-wide general AI agent (named Nan Xiao Yi). Data included 20,176 real-world logs (June–December 2025) and a cross-sectional survey of 152 transplant patients. We examined patterns of use over time, the types of questions raised, and the factors influencing patient behavior.

Results

Unlike Nan Xiao Yi, Doctor Xiao Yi remained active during OOH, with a peak at 4:00 AM (P < 0.001). The general agent handled admin tasks like appointments, while the specialist agent provided clinical support such as diet, symptoms, and medication. Survey found 60.5% of OOH use by transplant patients due to reluctance to disturb human doctors. Furthermore, 63.8% of transplant patients were satisfied with the specialist agent's responses, and 48% reported they would decide on further hospital treatment based on AI suggestions.

Conclusions

The specialist AI agent effectively fills the gap in medical and psychological services during OOH for transplant recipients. Based on the “dual-source knowledge base + GraphRAG + multi-agent framework” architecture, our specialist AI agent offers safe, reliable post-transplant care.

Keywords

Organ transplantation / Large language models (LLM) / Out-of-hours care / Artificial intelligence (AI) / Patient self-management / Graph-based retrieval-augmented generation (GraphRAG) / Mixed-methods study

References

[1] Jing W, Bi C, Fang Z, et al. . Neuropsychiatric sequelae after liver transplantation and their possible mechanism via the microbiota–gut–liver–brain axis. Biomed Pharmacother., 2023, 163 114855

[2] Cutting RB, Muscat DM, Patel P, et al. . Values, preferences, and risk tolerance of people waitlisted for a kidney transplant regarding potential deceased donor organ profiles: a systematic review. Transplantation., 2025, 109(6): e326-e339

[3] Olsen J, Wehberg S, Waldorff FB, et al. . General practice consultation patterns and patient factors predicting older patients’ use of out-of-hours services: a nationwide register-based cohort study. Br J Gen Pract., 2026, 76(763): e91-e99

[4] Bradshaw TJ, Tie X, Warner J, et al. . Large language models and large multimodal models in medical imaging: a primer for physicians. J Nucl Med., 2025, 66(2): 173-182

[5] Rao A, Pang M, Kim J, et al. . Assessing the utility of ChatGPT throughout the entire clinical workflow: development and usability study. J Med Internet Res., 2023, 25 e48659

[6] Xue E, Bracken-Clarke D, Iannantuono GM, et al. Utility of large language models for health care professionals and patients in navigating hematopoietic stem cell transplantation: comparison of the performance of ChatGPT-3.5, ChatGPT-4, and bard. J Med Internet Res. 2024;26:e54758.

[7] Stampe K, Kishik S, Müller SD. Mobile health in chronic disease management and patient empowerment: exploratory qualitative investigation into patient-physician consultations. J Med Internet Res., 2021, 23(6 e26991

[8] Velleca A, Welton M, Forsberg A. Moving Beyond Survival in Heart Transplantation. Can J Cardiol., 2026, 422): 417-420

[9] Koller P, Clement C, van Eijk A, et al. . Optimizing theranostics chatbots with context-augmented large language models. Theranostics., 2025, 15(12): 5693-5704

[10] Elangovan K, Ong JCL, Jin L, et al. . Development and evaluation of a lightweight large language model chatbot for medication enquiry. PLoS Digit Health., 2025, 4(9 e0000961

[11] Wang L, Li J, Zhuang B, et al. . Accuracy of large language models when answering clinical research questions: systematic review and network meta-analysis. J Med Internet Res., 2025, 27 e64486

[12] Schwarzkopf SC, Bereuter JP, Geissler ME, et al. . Postoperative complication management: How do large language models measure up to human expertise PLoS Digit Health., 2025, 48e0000933

[13] de Dumast L, Moore P, Snell KI, et al. . Trends in clinical workload in UK primary care 2005–2019: a retrospective cohort study. Br J Gen Pract., 2024, 74747): e659-e665

[14] Laymouna M, Ma Y, Lessard D, et al. . Roles, users, benefits, and limitations of chatbots in health care: rapid review. J Med Internet Res., 2024, 26 e56930

[15] Esbensen KL, DeCamp M, Crigger EJ, et al. . Ethical issues in organ transplantation: a position paper from the American college of physicians. Ann Intern Med., 2025, 178(12): 1772-1778

Funding

the National Outstanding Youth Physician(National High-level Medical Talent)

the National Natural Science Foundation of China(82570781)

the Key Project of Scientific and Technological Achievements Transformation of Zhongnan Hospital, Wuhan University(LCYFZD2024008)

the General Program of Hubei Provincial Health Science and Technology(WJ2025M062)

RIGHTS & PERMISSIONS

The Author(s)

Just Accepted

This article has successfully passed peer review and final editorial review, and will soon enter typesetting, proofreading and other publishing processes. The currently displayed version is the accepted final manuscript. The officially published version will be updated with format, DOI and citation information upon launch. We recommend that you pay attention to subsequent journal notifications and preferentially cite the officially published version. Thank you for your support and cooperation.