机器学习算法对急性胆管炎患者28天内死亡率的预测分析

A predictive analysis of 28-day mortality in patients with acute cholangitis using machine learning algorithms

摘要

目的：构建机器学习模型，评估并解释其对重症监护室（ICU）内急性胆管炎（AC）患者28 d内死亡率的预测效能。方法：回顾性采集了重症医学公共数据库MIMIC-IV（v3.1）收录的首次进入ICU治疗的AC患者数据，随机按照7∶3的比例分为训练集和验证集，采用最小绝对收缩和选择算子（LASSO）算法进行重要特征筛选，构建6种可监督机器学习预测模型，基于曲线下面积（AUC）确定最佳模型。分别利用受试者工作特征曲线和临床决策曲线评估最佳模型预测效能，沙普利可加性解释方法（SHAP）进行解释性分析。结果：共纳入555例AC患者，训练集388例、验证集167例。LASSO算法筛选得出5个重要特征。Catboost模型的AUC值在6种模型中达到最高，其在训练集和验证集中均表现出了良好的区分度，AUC值分别为0.850（95% CI：0.795~0.893）和0.839（95% CI：0.762~0.915）；临床决策曲线提示该模型具有良好的临床净获益概率。SHAP分析表明，筛选得到的5个特征对结局的预测重要性排序依次为急性生理评分系统Ⅲ、白蛋白、红细胞分布宽度、部分凝血活酶时间及碱性磷酸酶，而且全局解释和个体解释都具有临床合理性和适用性。结论：可解释性Catboost机器学习模型对AC患者28 d死亡率具备良好的预测效能，有望为ICU医师个体化决策提供参考。

Abstract

Objective: Developing machine learning algorithm models to evaluate and interpret their predictive efficacy for 28-day mortality in patients with acute cholangitis (AC) admitted to intensive care units (ICU). Methods: The data from ICU patients with AC were retrospectively collected from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database (v3.1). The patients were randomly divided into training and validation sets at a 7∶3 ratio. The Least Absolute Shrinkage and Selection Operator (LASSO) regression method was used for feature selection, and six supervised machine learning prediction models were constructed. The performance of the optimal predictive model was evaluated using receiver operating characteristic curves and clinical decision curves, with interpretability analysis conducted via the SHapley Additive exPlanations (SHAP) method. Results: A total of 555 AC patients was included, comprising 388 cases in the training set and 167 cases in the validation set, respectively. The LASSO algorithm identified five significant features. The Catboost model achieved the highest AUC value among the six models, demonstrating good discriminatory ability in both the training (AUC=0.850, 95% CI: 0.795~0.893) and validation (AUC=0.839, 95% CI: 0.762~0.915) sets. The clinical decision curve indicated that the model possesses a favorable net clinical benefit probability. SHAP analysis revealed that the importance of the five selected features ranked as follows: Acute Physiology Score Ⅲ (APS Ⅲ), albumin, red cell distribution width, partial thromboplastin time, and alkaline phosphatase. Both global and individual interpretations demonstrated clinical plausibility and applicability. Conclusion: The interpretable Catboost machine learning model developed in this study demonstrates strong predictive performance for 28-day mortality in AC patients, promising individualized decision-making for ICU physicians.

关键词

急性胆管炎 / 死亡 / 机器学习 / 预测模型 / 重症监护室 / 可解释性

Key words

Acute cholangitis / Mortality / Machine learning / Predictive model / Intensive care unit / Interpretability

参考文献

[1] Cozma MA , Găman MA , Srichawla BS , et al. Acute cholangitis: a state—of—the—art review[J]. Ann Med Surg (Lond), 2024, 86(8): 4560-4574. DOI: 10.1097/MS9.0000000000002169.

[2] Lavillegrand JR , Mercier—Des—Rochettes E , Baron E , et al. Acute cholangitis in intensive care units: Clinical, biological, microbiological spectrum and risk factors for mortality: a multicenter study[J]. Crit Care, 2021, 25(1): 49. DOI: 10.1186/s13054—021—03480—1.

[3] Yokoe M , Hata J , Takada T , et al. Tokyo guidelines 2018: Diagnostic criteria and severity grading of acute cholecystitis (with videos)[J]. J HBPS, 2018, 25(1): 41-54. DOI: 10.1002/jhbp.515.

[4] Sperna Weiland CJ , Busch CBE , Bhalla A , et al. Performance of diagnostic tools for acute cholangitis in patients with suspected biliary obstruction[J]. JHBPS, 2022, 29(4): 479-486. DOI: 10.1002/jhbp.1096.

[5] Haug CJ , Drazen JM . Artificial intelligence and machine learning in clinical medicine, 2023[J]. N Engl J Med, 2023, 388(13): 1201-1208. DOI: 10.1056/NEJMra2302038.

[6] 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. DOI: 10.1111/joim.12822.

[7] 牟鑫涛， 曹健强， 杨胜飚, 等 . 人工智能技术在现代普外科中的应用进展[J]. 中国现代普通外科进展， 2024， 27（5）： 337-342. DOI: 10.3969/j.issn.1009—9905.2024.05.001.

[8] Liu SW , Li P , Li XQ , et al. Recent advances in machine learning for precision diagnosis and treatment of esophageal disorders[J]. World J Gastroenterol, 2025, 31(23): 105076. DOI: 10.3748/wjg.v31.i23.105076.

[9] Ligero M , El Nahhas OSM , Aldea M , et al. Artificial intelligence—based biomarkers for treatment decisions in oncology[J]. Trends Cancer, 2025, 11(3): 232-244. DOI: 10.1016/j.trecan.2024.12.001.

[10] Peng S , Huang J , Liu X , et al. Interpretable machine learning for 28—day all—cause in—hospital mortality prediction in critically ill patients with heart failure combined with hypertension: a retrospective cohort study based on medical information mart for intensive care database—IV and eICU databases[J]. Front Cardiovasc Med, 2022, 9: 994359. DOI: 10.3389/fcvm.2022.994359.

[11] Han Y , Xie X , Qiu J , et al. Early prediction of sepsis associated encephalopathy in elderly ICU patients using machine learning models: a retrospective study based on the MIMIC—Ⅳ database[J]. Front Cell Infect Microbiol, 2025, 15: 1545979. DOI: 10.3389/fcimb.2025.1545979.

[12] Singer M , Deutschman CS , Seymour CW , et al. The third international consensus definitions for sepsis and septic shock (sepsis—3)[J]. JAMA, 2016, 315(8): 801-810. DOI: 10.1001/jama.2016.0287.

[13] Ponce—Bobadilla AV , Schmitt V , Maier CS , et al. Practical guide to SHAP analysis: explaining supervised machine learning model predictions in drug development[J]. Clin Transl Sci, 2024, 17(11): e70056. DOI: 10.1111/cts.70056.

[14] Sung SM , Kang YJ , Cho HJ , et al. Prediction of early neurological deterioration in acute minor ischemic stroke by machine learning algorithms[J]. Clin Neurol Neurosurg, 2020, 195: 105892. DOI: 10.1016/j.clineuro.2020.105892.

[15] Adams R , Henry KE , Sridharan A , et al. Prospective, multi—site study of patient outcomes after implementation of the TREWS machine learning—based early warning system for sepsis[J]. Nat Med, 2022, 28(7): 1455-1460. DOI: 10.1038/s41591—022—01894—0.

[16] Li Y , Wu Y , Gao Y , et al. Machine—learning based prediction of prognostic risk factors in patients with invasive candidiasis infection and bacterial bloodstream infection: a singled centered retrospective study[J]. BMC Infect Dis, 2022, 22(1): 150. DOI: 10.1186/s12879—022—07125—8.

[17] Park SW , Yeo NY , Kang S , et al. Early prediction of mortality for septic patients visiting emergency room based on explainable machine learning: A real—world multicenter study[J]. J Korean Med Sci, 2024, 39(5): e53. DOI: 10.3346/jkms.2024.39.e53.

[18] Zhou H , Liu L , Zhao Q , et al. Machine learning for the prediction of all—cause mortality in patients with sepsis—associated acute kidney injury during hospitalization[J]. Front Immunol, 2023, 14: 1140755. DOI: 10.3389/fimmu.2023.1140755.

[19] Xie W , Li Y , Meng X , et al. Machine learning prediction models and nomogram to predict the risk of in—hospital death for severe DKA: a clinical study based on MIMIC—Ⅳ, eICU databases, and a college hospital ICU[J]. Int J Med Inf, 2023, 174: 105049. DOI: 10.1016/j.ijmedinf.2023.105049.

[20] Chen Z , Li T , Guo S , et al. Machine learning—based in—hospital mortality risk prediction tool for intensive care unit patients with heart failure[J]. Front Cardiovasc Med, 2023, 10: 1119699. DOI: 10.3389/fcvm.2023.1119699.

[21] Pedarzani E , Fogangolo A , Baldi I , et al. Prioritizing patient selection in clinical trials: a machine learning algorithm for dynamic prediction of in—hospital mortality for ICU admitted patients using repeated measurement data[J]. J Clin Med, 2025, 14(2): 612. DOI: 10.3390/jcm14020612.

[22] Liu J , Song Y , Zhang D , Ji J . Sex differences in the association between red cell distribution width and 30—day mortality in critically ill patients with acute cholangitis: A retrospective cohort study[J]. BMC Gastroenterol, 2025, 26: 59. DOI: 10.1186/s12876—025—04549—9.

[23] Kalas MA , Chavez L , Leon M , et al. Abnormal liver enzymes: A review for clinicians[J]. World J Hepatol, 2021, 13(11): 1688-1698. DOI: 10.4254/wjh.v13.i11.1688.