Value analysis of predictive model of acute kidney injury in patients after cardiothoracic surgery based on random forest algorithm
-
摘要:
目的 构建基于随机森林算法的心胸外科术后患者急性肾损伤(acute kindey injury,AKI)的预测模型,并分析其预测价值。 方法 选取广元市第一人民医院2018年1月~2020年12月接受心胸外科手术治疗的212例患者为研究对象,采用随机数字表法按2:1的比例建立训练集和测试集。采用随机森林算法对心胸外科患者术后AKI的预测指标的重要性进行排序。根据袋外数据误差,赤池信息量准则和贝叶斯信息量准则对排序指标进行筛选并构建预测模型,多维标度法(multidimensional scaling,MDS)观察预测模型对心胸外科术后患者AKI的预测能力;采用内部验证法验证模型对心胸外科患者术后AKI的预测能力。 结果 212例患者中,148例未发生AKI的为未发生组,64例发生AKI为发生组;16 个指标根据平均准确度下降程度和平均基尼指数下降程度进行重要性排序。用袋外数据误差,赤池信息量准则和贝叶斯信息量准则筛选出术后中性粒细胞明胶酶相关脂质运载蛋白(NGAL)、术后金属蛋白酶组织抑制剂2(TIMP 2)、术后胰岛素样生长因子结合蛋白7(IGFBP7)及术后血浆肌酐(pCr)4个变量(P<0.01),并纳入模型。 通过 MDS 法观察到预测模型区分良好。 结论 基于术后NGAL、术后TIMP 2、术后IGFBP7及术后pCr建立的随机森林算法预测模型可用于心胸外科术后患者AKI发生的预测。 Abstract:AIM To construct a predictive model of acute kidney injury (AKI) for patients undergoing cardiothoracic surgery based on the random forest algorithm and analyze its predictive value. METHODS Two hundred and twelve patients in our hospital who underwent cardiothoracic surgery from January 2018 to December 2020 were selected as the research subjects. The random number table method was used to establish the training set and the test was set at a ratio of 2:1. Random forest algorithm was used to rank the importance of predictive indicators of AKI after cardiothoracic surgery. According to the out-of-bag data error, the Akaike information criterion and the Bayesian information criterion, the ranking indexes were screened and the predictive model was constructed. The multidimensional scaling method (MDS) was used to observe the predictive ability of the model after cardiothoracic surgery and internal verification method was adopted to validate the model’s ability to predict postoperative AKI in patients undergoing cardiothoracic surgery. RESULTS Among the 212 patients, AKI occurred in 64 cases (occurrence group) and did not occur in 148 cases (non-occurring group). The 16 indicators were ranked in importance according to the degree of decline in average accuracy and the degree of decline in average Gini index. Using the out-of-bag data error, Akaike information criterion and Bayesian information criterion, 4 variables were screened out, namely, postoperative NGAL, postoperative TIMP 2, postoperative IGFBP7 and postoperative pCr, and they were included in the model(P<0.01). It was observed that the prediction models were well differentiated by the MDS method. CONCLUSION The random forest algorithm prediction model based on postoperative NGAL, postoperative TIMP 2, postoperative IGFBP7 and postoperative pCr can be used to predict the occurrence of AKI in patients after cardiothoracic surgery. -
Key words:
- random forest algorithm /
- cardiothoracic surgery /
- acute kidney injury /
- prediction model
-
表 1 两组患者一般资料比较
项目 未发生组
(n=148)发生组
(n=64)年龄(岁) 65 ± 12 68 ± 12 男性 86(58) 40(62) BMI(kg/m2) 26.8 ± 2.8 26.4 ± 3.5 饮酒 62(42) 34(53) 吸烟 131(88) 59(92) 合并症 高血压 65(44) 37(58) 高脂血症 23(16) 11(17) 糖尿病 30(20) 16(25) 慢阻肺 19(13) 8(12) 术前平动脉压(mmHg) 88 ± 11 92 ± 10 a 术前左室射血分数(%) 59 ± 11 55 ± 9 a 术前pCr(mg/dL) 55 ± 14 50 ± 15 a 术前eGFR(ml/min/1.73m2) 88 ± 10 90 ± 13 术前白蛋白(g/L) 40 ± 4 38 ± 5 b 手术类型 冠状动脉旁路移植术 75(51) 32(50) 心脏瓣膜置换或修术 73(49) 32(50) CPB时间(min) 100 ± 30 109 ± 24 a 术中输血 96(65) 43(16) 术后24 h尿量(mL) 2135 ± 307 1932 ± 407 b 术后白细胞(×109/L) 10 ± 2 11 ± 2 b 术后NGAL(ug/L) 105 ± 15 113 ± 8 b 术后pCysC(mg/L) 1.6 ± 0.4 2.8 ± 0.6 b 术后pUrea(mmol/L) 4.3 ± 1.4 4.8 ± 1.6 a 术后pCr 45 ± 13 49 ± 13 a 术后血清L-FABP[μg/(g/Cr)] 13.3 ± 1.2 12.2 ± 2.5 b 术后KIM-1(mg/L) 1.3 ± 0.2 1.5 ± 0.3 b 术后TIMP 2(ng/L) 2.3 ± 0.7 4.2 ± 1.1 b 术后IGFB 7(ng/L) 0.2 ± 0.1 0.4 ± 0.1 b 表中计数资料均为[例数(%)],与未发生组比较,aP<0.05, bP<0.01 表 2 心胸外科术后患者AKI发生预测模型的变量筛选
项目 平均准确度下降程度 平均基尼指数下降程度 袋外数据误差(%) 赤池信息量 贝叶斯信息量 袋外数据误差(%) 赤池信息量 贝叶斯信息量 前2 28.59 164.45 173.50 28.57 164.45 173.50 前3 31.88 164.80 176.87 25.93 163.84 174.90 前4 31.22 162.96 177.84 26.58 164.15 173.24 前5 31.88 164.80 182.09 27.24 163.60 174.50 前6 26.58 164.50 185.62 25.26 161.94 182.87 前7 25.26 162.80 192.97 26.58 162.99 190.15 前8 25.92 161.25 194.44 25.26 159.17 189.46 -
[1] 中国生物医学工程学会体外循环分会. 2020年中国心外科手术和体外循环数据白皮书[J]. 中国体外循环杂志, 2021, 19(5): 257 – 260. doi: 10.13498/j.cnki.chin.j.ecc.2021.05.01 [2] 薛瑾虹, 胡淑芹, 陈 蕾, 等. 心脏外科手术后急性肾损伤患者的临床特征[J]. 肾脏病与透析肾移植杂志, 2019, 28(4): 324 – 329. doi: 10.3969/j.issn.1006-298X.2019.04.005 [3] 侯克龙, 李 静, 穆心苇. 心血管外科术后急性肾损伤的诊疗进展[J]. 中国心血管杂志, 2020, 25(1): 92 – 95. doi: 10.3969/j.issn.1007-5410.2020.01.022 [4] 高绪霞, 马立萍, 马涵英, 等. 4878例心脏外科手术后患者急性肾损伤发生率及其危险因素分析[J]. 中华肾脏病杂志, 2020, 36(5): 359 – 365. doi: 10.3760/cma.j.cn441217-20200103-00069 [5] 于 洋, 贾维坤. 心脏手术相关急性肾损伤的新型生物标志物研究进展[J]. 河北医药, 2020, 42(3): 457 – 462. doi: 10.3969/j.issn.1002-7386.2020.03.035 [6] Yuan SM. Acute kidney injury after cardiac surgery: risk factors and novel biomarkers[J]. Braz J Cardiovasc Surg, 2019, 34(3): 352 – 360. doi: 10.21470/1678-9741-2018-0212 [7] Doi K, Nishida O, Shigematsu T, et al. Japanese clinical practice guideline for acute kidney injury 2016 committee. the Japanese clinical practice guideline for acute kidney injury 2016[J]. Clin Exp Nephrol, 2018, 22(5): 985 – 1045. doi: 10.1007/s10157-018-1600-4 [8] 张芸楠, 贾 明, 潘 昱, 等. 心脏手术术后发生急性肾损伤的相关因素[J]. 中华医学杂志, 2020, 100(12): 928 – 932. doi: 10.3760/cma.j.cn112137-20190712-01555 [9] 龚海蓉, 雷 翀, 熊利泽. 心脏手术后急性肾损伤风险预测模型的建立: 文献分析[J]. 中华麻醉学杂志, 2020, 40(1): 18 – 26. doi: 10.3760/cma.j.issn.0254-1416.2020.01.005 [10] Lei L, Li LP, Zeng Z, et al. Value of urinary KIM-1 and NGAL combined with serum Cys C for predicting acute kidney injury secondary to decompensated cirrhosis[J]. Sci Rep, 2018, 8(1): 7962. doi: 10.1038/s41598-018-26226-6 [11] Cole OM, Tosif S, Shaw M, et al. Acute kidney injury and postoperative atrial fibrillation in patients undergoing cardiac surgery[J]. J Cardiothorac Vasc Anesth, 2020, 34(7): 1783 – 1790. doi: 10.1053/j.jvca.2019.12.048 [12] McIlroy DR, Farkas D, Matto M, et al. Neutrophil gelatinase-associated lipocalin combined with delta serum creatinine provides early risk stratification for adverse outcomes after cardiac surgery: a prospective observational study[J]. Crit Care Med, 2015, 43(5): 1043 – 1052. doi: 10.1097/CCM.0000000000000927 [13] Ho J, Reslerova M, Gali B, et al. Serum creatinine measurement immediately after cardiac surgery and prediction of acute kidney injury[J]. Am J Kidney Dis, 2012, 59(2): 196 – 201. doi: 10.1053/j.ajkd.2011.08.023 [14] Wang W, Li Z, Chen Y, et al. Prediction value of serum NGAL in the diagnosis and prognosis of experimental acute and dhronic kidney injuries[J]. Biomolecules, 2020, 10(7): 981. doi: 10.3390/biom10070981 [15] Albert C, Zapf A, Haase M, et al. Neutrophil gelatinase-associated lipocalin measured on clinical laboratory platforms for the prediction of acute kidney onjury and the associated need for dialysis therapy: a systematic review and meta-analysis[J]. Am J Kidney Dis, 2020, 76(6): 826 – 841,e1. doi: 10.1053/j.ajkd.2020.05.015 [16] Jia HM, Huang LF, Zheng Y, et al. Diagnostic value of urinary tissue inhibitor of metalloproteinase-2 and insulin-like growth factor binding protein 7 for acute kidney injury: a meta-analysis[J]. Crit Care, 2017, 21(1): 77 – 85. doi: 10.1186/s13054-017-1660-y [17] Wang X, Rojas-Quintero J, Wilder J, et al. Tissue inhibitor of metalloproteinase-1 promotes polymorphonuclear neutrophil (PMN) pericellular proteolysis by anchoring matrix metalloproteinase-8 and -9 to PMN surfaces[J]. J Immunol, 2019, 202(11): 3267 – 3281. doi: 10.4049/jimmunol.1801466 [18] Yan Y, Jiang L, Li M, et al. Levels of matrix metalloproteinase-9 and tissue inhibitor of metalloproteinase-1 are related to cardiopulmonary injury in fetal inflammatory response syndrome[J]. Clinics (Sao Paulo), 2020, 75: e2049. doi: 10.6061/clinics/2020/e2049 -