慢加急性肝衰竭患者90 d病死率的影响因素分析及机器学习应用

摘要/Abstract

摘要： 目的探究慢加急性肝衰竭(ACLF)患者 90 d病死率的影响因素,构建机器学习预测模型。方法本研究采用单中心回顾性队列研究设计,纳入 2020年6月1日至2024年5月31日北京地坛医院 534 例ACLF患者,按 7∶3 的比例分为训练集与测试集,经 LASSO回归和 10 折交叉验证筛选变量,用随机森林算法建模,通过 Kaplan-Meier曲线、时间依赖 ROC 曲线及 SHAP 值验证模型。结果凝血酶原活动度(PTA)、中性粒细胞、血磷、血氨、粒淋比(NLR)为独立危险因素。随机森林时间依赖 ROC 曲线模型预测 28 d、90 d预后的曲线下面积(AUC)分别为 0.89、0.88,验证集 AUC 分别为 0.88、0.82。经 surv_cutpoint 函数最大优度法确定,模型评分最优截点为 10.32,以此将患者分为高、低风险组,两组生存差异具有显著统计学意义(P<0.001)。与其他评分系统相比,森林图模型预测 28 d、90 d预后的 AUC(0.89、0.83)更优。SHAP 值可视化显示,各因素权重依次为:NLR＞血氨＞PTA＞中性粒细胞＞血磷;高风险组中,NLR、血氨、中性粒细胞计数、血磷的 SHAP 权重值分别增加 6.56、3.73、1.42、0.32,PTA 的 SHAP 权重值降低1.37,均与死亡风险升高相关。结论 PTA、中性粒细胞、血磷、血氨及NLR是影响ACLF患者短期预后的核心要素,机器学习在预后预测方面具有优势,可为临床实现精准干预提供全新手段。

关键词: 慢加急性肝衰竭, 90 d病死率, 影响因素, 机器学习, 预测模型

Abstract: Objective To explore the risk factors for 90-day mortality in patients with acute-on-chronic liver failure (ACLF) and construct a machine learning-based prediction model. Methods A single-center retrospective cohort study was conducted, including 534 ACLF patients admitted to Beijing Ditan Hospital from June 1, 2020, to May 31, 2024. The dataset was divided into training and test sets at a 7∶3 ratio. Variables were screened via Lasso regression and 10-fold cross-validation, and a random forest algorithm was used to build the model, which was validated by Kaplan-Meier curves, time-dependent ROC curves, and SHAP value analysis. Results Prothrombin activity (PTA), neutrophils, serum phosphorus, blood ammonia, and neutrophil-to-lymphocyte ratio (NLR) were identified as independent risk factors. The area under the curve (AUC) of the random forest time-dependent ROC curve model for predicting 28-day and 90-day prognosis was 0.89 and 0.88, respectively, with the AUC of the validation set being 0.88 and 0.82, respectively. Determined by the maximum goodness-of-fit method via the surv_cutpoint function, the optimal cut-off value for the model score was 10.32. Patients were divided into high-risk and low-risk groups based on this value, and the survival difference between the two groups was statistically significant (P<0.001). Compared with other scoring systems, the forest plot model showed better AUC values (0.89, 0.83) in predicting 28-day and 90-day prognosis. SHAP value visualization indicated that the weight of each factor was in the order of: NLR > blood ammonia > PTA > neutrophils > serum phosphorus; in the high-risk group, the SHAP weight values of NLR, blood ammonia, neutrophil count, and serum phosphorus increased by 6.56, 3.73, 1.42, and 0.32, respectively, while the SHAP weight value of PTA decreased by 1.37, all of which were associated with an increased risk of death. Conclusion The study confirms that PTA, neutrophils, serum phosphorus, blood ammonia, and NLR are core factors affecting the short-term prognosis of ACLF patients. It also verifies the advantages of machine learning in prognostic prediction, providing a new means for precise clinical intervention.

Key words: Acute-on-chronic liver failure, 90-day mortality, Risk factors, Machine learning, Prediction model

郭贺冰, 刘景院, 万钢, 李昂. 慢加急性肝衰竭患者90 d病死率的影响因素分析及机器学习应用[J]. 肝脏, 2026, 31(4): 569-577.

GUO He-bing, LIU Jing-yuan, WAN Gang, LI Ang. Analysis of risk factors for 90-day mortality in patients with acute-on-chronic liver failure and machine learning application[J]. Chinese Hepatolgy, 2026, 31(4): 569-577.

/ 推荐

参考文献

[1] Sarin S K, Choudhury A, Sharma M K, et al. Acute-on-chronic liver failure: consensus recommendations of the Asian Pacific association for the study of the liver (APASL): an update[J].Hepatol Int, 2019, 13(4):353-390.
[2] 李乐辉, 吴国平, 林润.血清Trp联合血氨检测提高急诊ICU肝性脑病早期诊断灵敏度的临床价值[J].肝脏, 2019, 24(4):415-417.
[3] Chetwood J D, Sabih A H, Chan K, et al.Epidemiology, characteristics, and outcomes of patients with acute-on-chronic liver failure in Australia[J]. J Gastroenterol Hepatol, 2023, 38(8):1325-1332.
[4] 汤影子, 夏杰, 刘慧敏, 等.人工智能在慢性肝病领域中的应用研究进展[J].解放军医学杂志, 2022, 47(8):845-850.
[5] 郑琰莉, 苏文星, 宋元涛.机器学习在心血管疾病中应用的研究进展[J].中西医结合心脑血管病杂志, 2024, 22(10):1813-1817.
[6] 杨凯涵, 王旭亚, 冯晓彬, 等.肿瘤数智诊疗研究的现状及前景[J].中国肿瘤临床, 2023, 50(16):855-860.
[7] 涂颖, 李雪, 陈美娟, 等.血小板计数及相关评分模型对 HBV 相关慢加急性肝衰竭预后的预测价值[J].临床肝胆病杂志, 2023, 39(6):1308-1312.
[8] Sarin S K, Kumar A, Almeida J A, et al. Acute-on-chronic liver failure: consensus recommendations of the Asian Pacific Association for the study of the liver (APASL)[J]. Hepatol Int, 2009, 3(1):269-282.
[9] Amin A, Mookerjee R P.Acute-on-chronic liver failure: definition, prognosis and management[J]. Frontline Gastroenterol, 2019, 11(6):458-467.
[10] Zhang Y M, Li L J.State of the art-Artificial liver in China[J]. Artificial Organs, 2019, 43(4):336-341.
[11] Li P, Liang X, Luo J, et al.Predicting the survival benefit of liver transplantation in HBV-related acute-on-chronic liver failure: an observational cohort study[J].Lancet Reg Health West Pac, 2022, 32(100638):1-14.
[12] 赖曼, 徐曼曼, 王鑫,等.慢加急性肝衰竭肝移植预后评估[J]. 器官移植,2025, 16(3):482-488.
[13] Ge J, Digitale J C, Fenton C, et al.Predicting post-liver transplant outcomes in patients with acute-onchronic liver failure using expert-augmented machine learning[J]. Am J Transplant 2023, 23(12):1908-1921.
[14] 郭九麟, 李垣锟, 丁仁鑫,等.基于机器学习的急性肝衰竭死亡风险评估研究[J].中国数字医学, 2025, 20(3):20-25.
[15] 陈小琴, 林志刚.血清PCT,CRP,PTA联合MELD评分系统预测ACLF患者短期预后的价值[J].医学理论与实践, 2021, 34(23):4164-4166.
[16] 蒋雯, 郑泽美, 李军.凝血、纤溶与抗凝指标在慢性肝衰竭患者中的变化及相关性[J].检验医学与临床, 2019, 16(22):3340-3343.
[17] Baquerizo A, Anselmo D, Shackleton C, et al.Phosphorus ans an early predictive factor in patients with acute liver failure[J]. Transplantation, 2003, 75(12):2007-2014.
[18] Quirós-Tejeira R E, Molina R A, Katzir L, et al. Resolution of hypophosphatemia is associated with a recovery of hepatic function in children with fulminant hepatic failure[J].Gastroenterology, 2005, 18(9):1061-1066.
[19] 童晶晶, 穆秀颖, 许祥, 等.血氨与HBV相关慢加急性肝衰竭患者90天预后的关系[J].临床肝胆病杂志,2019, 35(6):1304-1307.
[20] Lattanzi B, D’Ambrosio D, Merli M. Hepatic encephalopathy and sarcopenia: two faces of the same metabolic alteration[J]. J Clin Exp Hepatol, 2019, 9(1):125-130.
[21] Albillos A, Lario M, Álvarez-Mon M.Cirrhosis-associated immune dysfunction: distinctive features and clinical relevance[J]. J Hepatol , 2014, 61(6):1385-1396.
[22] Choudhary N S, Saigal S, Saraf N, et al. Prognostic role of ammonia in cirrhotic patients[J]. Hepatology, 2019, 70(3):982-994.
[23] Patwardhan V R, Jiang Z G, Risech-Neiman Y, et al.Serum ammonia in associated with transplant-free survival in hospitalized patients with acutely decompensated cirrhosis[J]. J Clin Gastroenterol, 2016, 50(4):345-350.
[24] Dasarathy S, Merli M.Sarcopenia from mechanism to diagnosis and treatment in liver disease[J].J Hepatol ,2016, 65(6):1232-1244.
[25] Jalan R, Saliba F, Pavesi M, et al.Development and validation of a prognostic score to predict mortality in patients with acute-on-chronic liver failure[J]. J Hepatol, 2014, 61(5):1038-1047.
[26] Piano S, Brocca A, Mareso S, et al.Infections complicating cirrhosis[J].Liver Int,2018,38(Suppl. 1):126-131.
[27] Bernsmeier C, Singanayagam A, Patel V C, et al.Immunotherapy in the treatment and prevention of infection in acute-on-chronic liver failure[J]. Immunotherapy,2015, 7(6):641-654.
[28] Siddiqui M S, Stravitz R T.Intensive care unit management of patients with liver failure[J]. Clin Liver Dis, 2014, 18(4):957-978.
[29] Arvaniti V, D′Amico G, Fede G, et al.Infections in patients with cirrhosis increase mortality four-fold and should be used in determining prognosis[J]. Gastroenterology, 2010, 139(4):1246-1256.
[30] Clària J, Stauber R E, Coenraad M J, et al.Systemic inflammation in decompensated cirrhosis: characterization and role in acute-on-chronic liver failure[J]. Hepatology, 2016, 64(4):1249-1264.
[31] Bernsmeier C, Pop O T, Singanayagam A,et al.Patients with acute-on-chronic liver failure have increased numbers of regulatory immune cells expressing the receptor tyrosine kinase MERTK[J]. Gastroenterology, 2015, 148(3):603-615.
[32] Albillos A, Martin-Mateos R, van der Merwe S, et al.Cirrhosis-associated immune dysfunction[J].Nat Rev Gastroenterol Hepatol, 2022, 19(2):112-134.
[33] 曹桂花, 许荣, 薛明涛, 等.中性粒细胞百分比与高龄感染患者死亡率相关性分析[J].中华老年多器官疾病杂志, 2018, 17(1):39-42.
[34] Weiss E, de la Grange P, Defaye M, et al.Characterization of blood immune cells in patients with decompensated cirrhosis including ACLF[J]. Front Immunol, 2021, 5(11):1-14.
[35] Angus D C, van der Poll T. Severe sepsis and septic shock[J]. N Engl J Med, 2013, 369(8):840-851.
[36] Vidal A C, Howard L E, de Hoedt A,et al.Neutrophil, lymphocyte and platelet counts, and risk of prostate cancer outcomes in white and black men: results from the SEARCH database[J].Cancer Causes Control, 2018, 29(6):581-588.
[37] Hotchkiss R S, Opal S M. Activating immunity to fight a foe—a new path[J].N Engl J Med,2020, 382(13):1270-1272.
[38] Zhang X, Tan Y, Ling Y, et al.Viral and host factors related to the clinical outcome of COVID-19[J]. Nature, 2020, 583(7816):437-440.
[39] Rueschenbaum S, Ciesek S, Queck A, et al.Dysregulated adaptive immunity is an early event in liver cirrhosis preceding acute-on-chronic liver failure[J]. Front Immunol, 2021, 26(11):1-11.
[40] Moreau N, Wittebole X, Fleury Y, et al.Neutrophil-to-lymphocyte ratio predicts death in acute-on-chronic liver failure patients admitted to the intensive care unit: a retrospective cohort study[J]. Shock, 2017, 49(4):385-392.
[41] Cai J, Wang K, Han T, et al. Evaluation of prognostic values of inflammation-based makers in patients with HBV-related acute-on-chronic liver failure[J]. Medicine, 2018, 97(46):1-10.
[42] Jiang J, Liu R, Yu X, et al.The neutrophil-lymphocyte count ratio as a diagnostic marker for bacteraemia: a systematic review and meta-analysis[J]. Am J Emerg Med, 2019, 37(8):1482-1489.
[43] Jindal A, Bhadoria A S, Maiwall R, et al.Evaluation of acute kidney injury and its response to terlipressin in patients with acute-on-chronic liver failure[J]. Liver Int, 2016, 36(1):59-67.