目的 探讨大语言模型（LLMs）在简化冠状动脉旁路移植术（CABG）风险评估的应用可行性和潜力,以及模型的可解释性和临床适用性。方法 基于两家心脏中心1 372例CABG患者临床数据,采用OpenAI GPT-o3模型对每例患者术前资料进行语义精简与临床映射,通过少样本提示方式重构评估框架,构建简化手术风险评估模型,并将其与欧洲心血管手术风险评估系统Ⅱ（EuroSCORE Ⅱ）评分模型进行一致性、准确性等比较。结果 LLMs可有效识别手术高相关性风险因子,自动生成结构化手术风险评估提示,所构建的模型在现有数据库中受试者工作特征曲线下面积、灵敏度、特异度等优于EuroSCORE Ⅱ评分模型,并在高风险患者识别上表现突出。结论 LLMs具备辅助构建、简化手术风险评估模型能力,具有临床可适用性和替代现有心脏手术风险评估模型的潜力,可为实现个体化手术风险评估与智能辅助决策提供新的思路。
Objective To explore the feasibility and potential of large language models （LLMs） in streamlining risk assessment process of coronary artery bypass grafting （CABG）, with a focus on interpretability and clinic applicability. Methods Using clinical data from 1 372 CABG patients at two cardiac centers, the OpenAI GPT-o3 model was used to process preoperative information to create a simplified risk assessment model via few-shot prompting. This LLM-based model's performance was compared to the European System for Cardiac Operative Risk Evaluation Ⅱ （EuroSCOREⅡ） for consistency and accuracy. Results LLMs excelled at detecting high surgical risks and generating structured risk assessments, surpassing EuroSCOREⅡ in area under the curve （AUC） of receiver operating characteristic, sensitivity, and specificity, thus better identifying high-risk patients. Conclusion LLMs can create and simplify surgical risk assessment models with strong clinical use, potentially complementing or replacing current cardiac surgery risk assessment tools and offering new options for personalized risk evaluation and decision support in clinical settings.
[1] Zhu L, Li D, Zhang X, et al.Comparative efficacy on outcomes of C-CABG, OPCAB, and ONBEAT in coronary heart disease: a systematic review and network meta-analysis of randomized controlled trials[J]. Int J Surg,2023,109(12):4263-4272.
[2] Zhang T, Yang X, Zhuang Q, et al.Study on myocardial infarction based on nationwide inpatient sample database: a bibliometric analysis from 2000 to 2022[J]. Emerg Crit Care Med,2024,4(1):22-27.
[3] Mastroiacovo G, Bonomi A, Ludergnani M, et al. Is EuroSCORE Ⅱ still a reliable predictor for cardiac surgery mortality in2022? A retrospective study study[J]. Eur J Cardiothorac Surg,2022,64(3):ezad294.
[4] Bedi S, Liu Y, Orr-Ewing L et al. Testing and evaluation of health care applications of large language models: a systematic review[J]. JAMA,2025,333(4):319-328.
[5] Allyn J, Allou N, Augustin P et al. A comparison of a machine learning model with EuroSCORE Ⅱ in predicting mortality after elective cardiac surgery: a decision curve analysis[J]. PLoS One,2017,12(1):e0169772.
[6] Hijazi E. Enhancing mortality prediction after coronary artery bypass graft: a machine learning approach utilizing EuroScore[J]. Future Sci OA,2024,10(1):FSO959.
[7] Doenst T, Haverich A, Serruys P, et al.PCI and CABG for treating stable coronary artery disease: JACC review topic of the week[J]. J Am Coll Cardiol,2019,73(8):964-976.
[8] Schafer A, Alasnag M, Giacoppo D, et al.High-risk percutaneous coronary intervention in patients with reduced left ventricular ejection fraction deemed not suitable for surgical revascularisation. A clinical consensus statement from the European Association of Percutaneous Cardiovascular Interventions (EAPCI) in collaboration with the ESC Working Group on Cardiovascular Surgery[J]. EuroIntervention,2025,21(1):22-34.
[9] Nashef SA, Sharples LD, Roques F, et al.EuroSCORE Ⅱ and the art and science of risk modelling[J]. Eur J Cardiothorac Surg,2013,43(4):695-696.
[10] Di Dedda U, Pelissero G, Agnelli B, et al.Accuracy, calibration and clinical performance of the new EuroSCORE Ⅱ risk stratification system[J]. Eur J Cardiothorac Surg,2013,43(1):27-32.
[11] Neumann FJ, Sousa-Uva M, Ahlsson A, et al.2018 ESC/EACTS guidelines on myocardial revascularization[J]. Eur Heart J,2019,40(2):87-165.
[12] Di Bacco L, D'Alonzo M, Baudo M, et al. Reliability of EuroSCORE Ⅱ on prediction of thirty-day mortality and long-term results in patients treated with sutureless valves[J]. J Clin Med,2024,13(13):3986.
[13] Birlik AB, Tozan H, Kose KB.Machine learning-based hybrid risk estimation system (ERES) in cardiac surgery: supplementary insights from the ASA score analysis[J]. PLoS Digit Health,2025,4(6):e0000889.
[14] Fang SY, Chen JW, Chou HW, et al.Validation of the European system for cardiac operative risk evaluation Ⅱ in a large Taiwan cardiac surgical centre[J]. J Formos Med Assoc, 2023,122(12):1265-1273.
[15] Arfat Y, Mittone G, Esposito R, et al.Machine learning for cardiology[J]. Minerva Cardiol Angiol,2022,70(1):75-91.
[16] Handelman GS, Kok HK, Chandra RV, et al.eDoctor: machine learning and the future of medicine[J]. J Intern Med, 2018,284(6):603-619.
[17] Jia T, Xu K, Bai Y, et al.Machine-learning predictions for acute kidney injuries after coronary artery bypass grafting: a real-life muticenter retrospective cohort study[J]. BMC Med Inform Decis Mak,2023,23(1):270.
