北京生物医学工程

决策树算法应用于MIMIC-III数据库的ICU患者急性肾损伤预测研究

Decision tree algorithm applied to MIMIC-III database for the prediction of acute kidney injury in ICU patients

作者：高文鹏吕海金周琅郭圣文

单位：华南理工大学材料科学与工程学院生物医学工程系 (广州 510006) 中山大学附属第三医院外科ICU (广州510630)  华南理工大学自动化科学与工程学院(广州510640) 通信作者:郭圣文。 E-mail: shwguo@scut.edu.cn

关键词：急性肾损伤;重症监护室;机器学习;风险预测;重要特征

分类号：R318

出版年·卷·期（页码）：2021·40·6（609-617）

摘要：

目的急性肾损伤(acute kidney injury,AKI)是重症监护病房(intensive care unit,ICU)最常见的并发症和致死因素之一。准确预测具AKI风险的患者，明确与AKI发生相关的关键因素，可为临床决策与风险患者干预提供有效指导。方法采用公开的重症监护室数据库MIMIC-III,提取30 020例患者记录(包括AKI患者17 222名，Non-AKI患者12798名)，收集其住ICU期间基本信息、生理生化指标、药物使用、合并症等临床信息。将患者按4 ： 1比例随机划分训练集和独立测试集，应用逻辑回归、随机森林与LightGBM 3种机器学习方法，分别建立24 h、48 h与72 h 3个时间点的AKI预测模型，采用十折交叉验证法,对各种模型进行训练与测试,预测患者是否发生AKI,并获取重要特征。此外，利用24 h预测模型,在一周时间窗口内对ICU患者进行每隔24h预测。结果3种学习模型中,LightGBM性能最优，其24 h、48 h和72 h模型预测AKI的受试者工作特征曲线(receiver operator characteristic curve, ROC 曲线)下面积(area under curve, AUG)值分别为 0. 90.0. 88.0. 87,Fl 值分别为 0.91.0. 88,0. 86,在每隔24 h预测时，提前1 d、2 d和3 d预测AKI的成功率分别为89%、83%、80%。已住院时长、体质量、白蛋白、收缩压、碳酸氢盐、葡萄糖、白细胞计数、体温、舒张压、血尿素氮等是预测ICU患者AKI的重要特征,仅使用24个重要特征，模型仍能取得良好的预测性能。结论基于ICU患者的基本信息、生理生化指标、药物使用及合并症等临床信息，应用机器学习模型，可对其是否发生AKI进行多时间点的有效预测，并明确其关键风险因素。

Objective Acute kidney injury ( AKI) is one of the most common complications and fatal factors in intensive care unit (ICU). Accurate prediction of AKI risk and identification of key factors related to AKI can provide effective guidance for clinical decision-making and intervention for patients with AKI risk. Methods A total of 30020 patients in ICU (including 17 222 AKI patients and 12 798 Non-AKI patients) were selected from the public database MIMIC-III in this study, and basic information, physiological and biochemical indicators, drug use, and comorbidity during their stay in ICU were collected. All patients were randomly divided into training sets and independent testing sets according to the ratio of 4 ： 1, and logistic regression, random forest, and LightGBM were applied to construct models for AKI predication in three time points including 24 h, 48 h and 72 h, respectively. The 10-fold cross validation was used to train and validate various models to predict the occurrence of AKI,and obtain important features. Furthermore,24 h prediction models were used to predict AKI every 24 h during the 7-day window. Results LightGBM achieved the best performance with AUC values of 0. 90,0. 88,0. 87 for 24 h,48 h,and 72 h prediction,respectively,and Fl values were 0. 91,0. 88,and 0. 86. In prediction of every 24 h,the success rates of identifying AKI patients were 89% ,83%,and 80% in one day,two days and three days in advance, respectively. It was found that the length of stay in ICU, body weight, albumin, systolic blood pressure, bicarbonate, glucose, white blood cell count, body temperature, diastolic blood pressure and blood urea nitrogen played vital roles in predicting AKI for ICU patients. Using only 24 important features, the models could still achieve prominent prediction performance. Conclusions Based on basic information, physiological and biochemical indicators, drug use, and comorbidity, machine learning methods can be adopted to effectively predict AKI risk for ICU patients at several time points, and determine the dominant factors relative to AKI.

参考文献：

[1 ] Lameire NH, Bagga A, Cruz D, et al. Acute kidney injury: an increasing global concern [J]. Lancet, 2013, 382 ( 9887 ):170-179.

[2 ] Bagshaw SM, George C, Gibney RTN, et al. A multi-center evaluation of early acute kidney injury in critically ill trauma patients[J]. Renal Failure,2008,30(6) :581 -589.

[3 ] Mizuno T, Sato W, Ishikawa K, et aL KDIGO ( kidney disease: improving global outcomes) criteria could be a useful outcome predictor of cisplatin-induced acute kidney injury[J]. Oncology, 2012,82(6):354-359.

[4 ] Eriksson M, Brattstrom 0, Martensson J, et al. Acute kidney injury following severe trauma: risk factors and long-term outcome [J]. Journal of Trauma and Acute Care Surgery, 2015,79(3): 407-412.

[5 ] Ostermarm M, Joannidis M. Acute kidney injury 2016： diagnosis and diagnostic workup[J]. Critical Care,2016,20:299.

[6 ] Panagidis D, Nanas S, Kokkoris S. Biomarkers of acute kidney injury in a mixed ICU population, a narrative review [J]. Health & Research Journal, 2019,5(4):150.

[7 ] Rojas JC, Carey KA, Edelson DP ,et al. Predicting intensive care unit readmission with machine learning using electronic health record data[J]. Annals of the American Thoracic Society ,2018, 15(7):846-853.

[8 ] Haines RW, Lin SP, Hewson R, et al. Acute kidney injury in trauma patients admitted to critical care: development and validation of a diagnostic prediction model [J]. Scientific Reports ,2018,8:3665.

[9 ] Malhotra R, Kashani KB, Macedo E, et al. A risk prediction score for acute kidney injury in the intensive care unit[J]. Nephrology Dialysis Transplantation,2017,32(5) :814—822.

[10]?李千惠.基于机器学习的急性肾损伤预测及临床应用优化 [D].北京:北京交通大学,2019.

Li QH. Prediction of acute kidney injury and clinical application optimization based on machine learning [D]. Beijing； Beijing Jiaotong University ,2019.

[11]?张渊，冯聪，李开源，等.ICU患者急性肾损伤发生风险的 LightGBM预测模型[J].解放军医学院学报,2019,40(4): 316-320.

Zhang Y, Feng C, Li KY, et al. LightGBM model fbr predicting acute kidney injury risk in ICU patients [J]. Academic Journal of Chinese PLA Medical School,2019,40(4):316-320.

[12]?Zimmerman LP,Reyfinan PA,Smith A,et aL Early prediction of acute kidney injury following ICU admission using a multivariate panel of physiological measurements [J]. BMC Medical Informatics and Decision Making,2019,19(Suppl 1) : 16.

[13]?Johnson A, Pollard TJ, Shen L, et al. MIMIC - III, a freely accessible critical care database [J] . Scientific Data, 2016 ( 3 )： 160035.

[14]?Peng C, Waitman LR, Yong H, et al. Predicting inpatient acute kidney injury over different time horizons: How early and accurate? [J]. AMIA... Annual Symposium Proceedings/ AMIA Symposium, 2018,2017 ( 2017 ):565—574.

[15]?van Houwelingen S. Ridge estimators in logistic regression [J]. Journal of the Royal Statistical Society Series C, 1992,41 ( 1):191-201.

[16]?Criminisi A, Shotton J, Konukoglu E. Decision forests: a unified framework for classification, regression, density estimation, manifold learning and semi-supervised learning[J]. Foundations and Trends in Computer Graphics and Vision, 2011,7 ( 2- 3): 81-227.

[17]?Ke G, Meng Q, Finley T, et al. LightGBM: a highly efficient gradient boosting decision tree [C] //NIPS，17: the 31st International Conference on Neural Information Processing Systems. Long Beach, USA: NIPS ,2017,3149-3157.

[18]?Mallet ML. Pathophysiology of accidental hypothermia[J]. QJM, 2002,95(12) :775-785.

[19]?M6garbane B, Axler 0, Chary I, et al. Hypothermia with indoor occurrence is associated with a worse outcomef J]. Intensive Care Medicine,2000,26( 12):1843-1849.

[20]?Han SS, Ahn SY, Ryu J, et al. U -shape relationship of white blood cells with acute kidney injury and mortality in critically ill patients [ J ]. The Tohoku Journal of Experimental Medicine, 2014,232(3):177-185.

[21 ] Gujadhur A, Tiruvoipati R, Cole E, et al. Serum bicarbonate may independently predict acute kidney injury in critically ill patients: an observational study [ J ]. World Journal of Critical Care Medicine,2015,4(1) :71-76.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录