切换至 "中华医学电子期刊资源库"
高级检索
中华危重症医学杂志(电子版)

中华危重症医学杂志(电子版) ›› 2023, Vol. 16 ›› Issue (03) : 198 -206. doi: 10.3877/cma.j.issn.1674-6880.2023.03.004

论著

脓毒症继发急性肾损伤患者死亡风险预测模型构建及评估
莫小乔, 胡喆莹, 廖冬花, 谢天†()   
  1. 200092　上海，上海交通大学医学院附属新华医院手术室
    200092　上海，上海交通大学医学院附属新华医院小儿血液肿瘤科
    200011　上海，上海交通大学医学院附属第九人民医院普外科
  • 收稿日期:2022-09-20 出版日期:2023-06-30
  • 通信作者: 谢天

Establishment and evaluation of a predictive model for death risk in patients with acute kidney injury secondary to sepsis

Xiaoqiao Mo, Zheying Hu, Donghua Liao, Tian Xie†()   

  1. Department of Operative Room, Xinhua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai 200092, China
    Department of Pediatric Hematology Oncology, Xinhua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai 200092, China
    Department of General Surgery, Shanghai Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200011, China
  • Received:2022-09-20 Published:2023-06-30
  • Corresponding author: Tian Xie
全文 ( ) 下载PDF ( ) 导出引用
引用本文：

莫小乔, 胡喆莹, 廖冬花, 谢天. 脓毒症继发急性肾损伤患者死亡风险预测模型构建及评估[J]. 中华危重症医学杂志(电子版), 2023, 16(03): 198-206.

Xiaoqiao Mo, Zheying Hu, Donghua Liao, Tian Xie. Establishment and evaluation of a predictive model for death risk in patients with acute kidney injury secondary to sepsis[J]. Chinese Journal of Critical Care Medicine(Electronic Edition), 2023, 16(03): 198-206.

目的

建立评估脓毒症继发急性肾损伤（AKI）患者28 d死亡风险的预测模型。

方法

对Dryad免费开放数据库中AKI行连续性肾脏替代疗法（CRRT）的患者数据进行二次分析，应用Lasso回归筛选预测变量，并通过logistic回归分析脓毒症继发AKI患者28 d死亡的危险因素。根据预测因素应用R语言软件建立列线图预测模型，并采用受试者工作特征（ROC）曲线、校准曲线和决策曲线分析（DCA）探讨列线图预测模型与现有预测模型的效能及临床实用价值。最后应用Kaplan-Meier曲线分析列线图预测模型下不同危险分层组28 d及90 d的生存情况。

结果

共纳入798例脓毒症继发AKI患者，采用Lasso回归共筛选出包括白蛋白、24 h磷酸根离子、序贯器官衰竭估计（SOFA）评分、血肌酐、平均动脉压、尿素氮肌酐比、收缩压和2 h尿量在内的8个预测因素。Logistic回归分析显示预测因素中白蛋白[比值比（OR）= 0.589，95%置信区间（CI）（0.393，0.883），P = 0.010]、24 h磷酸根离子浓度[OR = 1.406，95%CI（1.225，1.613），P < 0.001]、SOFA评分[OR = 1.234，95%CI（1.152，1.321），P < 0.001]、血肌酐[OR = 0.773，95%CI（0.658，0.908），P = 0.002]、尿素氮肌酐比[OR = 1.017，95%CI（1.001，1.034），P = 0.037]、收缩压[OR = 0.982，95%CI（0.967，0.998），P = 0.023]和2 h尿量[OR = 0.997，95%CI（0.995，0.999），P = 0.011]为脓毒症继发AKI患者28 d死亡的独立影响因素。使用8个预测因素建立预测模型并通过列线图展示。列线图预测模型、Jung模型及APACHEⅡ评分ROC曲线下面积（0.839、0.809、0.618）、校准曲线拟合优度及DCA曲线高阈值概率范围（0.25 ~ 1.00、0.35 ~ 1.00、0.00 ~ 0.75）等综合模型评估上列线图模型较其他模型处于优势。基于列线图预测模型的3个危险分层组在28 d及90 d的总体生存时间分布比较差异均有统计学意义（χ2 = 178.847、180.665，P均< 0.001），且组间两两比较差异均有统计学意义（P均< 0.001）。

结论

列线图预测模型对脓毒症患者继发AKI 28 d死亡风险具有较好的预测效能，相较现有的预测工具，该模型变量少，性能优，且可实现个体化、可视化、图形化预测，可对临床早期识别并尽早采取措施改善患者病情转归提供一定的帮助。

关键词: 脓毒症, 急性肾损伤, 死亡风险, 预测模型
Objective

To establish a nomogram model for prediction of 28-day death in patients with acute kidney injury (AKI) secondary to sepsis.

Methods

This study was a secondary analysis on the data of AKI patients undergoing continuous renal replacement therapy in the Dryad Free Open Database. The predictive factors were selected by Lasso regression, and the risk factors for 28-day death were analyzed by logistic regression. A nomogram predictive model was established using R language software based on the predictive factors. The efficacy and clinical value of the nomogram predictive model were evaluated using the receiver operating characteristic (ROC) curves, calibration curves and decision curve analysis (DCA) and were compared with existing models. Kaplan-Meier curves were used to assess the survival of different stratification groups at 28 days and 90 days.

Results

A total of 798 patients with AKI secondary to sepsis were enrolled, and eight predictive factors were selected by the Lasso regression, including albumin, 24-hour phosphate ion concentration, sequential organ failure assessment (SOFA) score, serum creatinine, mean arterial pressure, urea nitrogen-creatinine ratio, systolic blood pressure and 2-hour urine output. Logistic regression analysis showed that albumin [odds ratio (OR) = 0.589, 95% confidence interval (CI) (0.393, 0.883), P = 0.010], 24-hour phosphate ion concentration [OR = 1.406, 95%CI (1.225, 1.613), P < 0.001], SOFA score [OR = 1.234, 95%CI (1.152, 1.321), P < 0.001], serum creatinine [OR = 0.773, 95%CI (0.658, 0.908), P = 0.002], urea nitrogen-creatinine ratio [OR = 1.017, 95%CI (1.001, 1.034), P = 0.037], systolic blood pressure [OR = 0.982, 95%CI (0.967, 0.998), P = 0.023], and 2-hour urine output [OR = 0.997, 95%CI (0.995, 0.999), P = 0.011] were independent risk factors for 28-day death in patients with AKI secondary to sepsis. A nomogram predictive model was established using the eight predictive factors. The area under ROC curves (0.839, 0.809, 0.618), goodness-of-fit of calibration curves and high threshold probability range of DCA curves (0.25-1.00, 0.35-1.00, 0.00-0.75) were performed on the nomogram model, Jung model and acute physiology and chronic health evaluation II model, and the nomogram model was superior to the other two models. The difference of overall survival among three stratification groups at 28 days and 90 days based on the nomogram predictive model was statistically significant (χ2 = 178.847, 180.665; both P < 0.001), and there were also significant differences between groups (all P < 0.001).

Conclusions

The nomogram predictive model shows good predictive efficacy on the risk of 28-day death in patients with AKI secondary to sepsis. Compared with the existing predictive tools, this model relies on fewer variables and shows better performance. Besides, the nomogram model has advantages in individualized, visualized and graphical prediction, which can provide assistance for early identification and early treatment to improve the outcomes of patients.

Key words: Sepsis, Acute kidney injury, Death risk, Predictive model
表1 脓毒症继发AKI患者一般资料情况比较[M（P25，P75）]
组别 例数 年龄（岁） 性别（例，男/女） 体质量指数（kg/m2） 收缩压（mmHg， ± s） 舒张压（mmHg， ± s） 平均动脉压（mmHg， ± s） Charlson合并症指数（ ± s） APACHEⅡ评分（分， ± s） SOFA评分（分， ± s） 2 h尿量（mL）
28 d生存组 300 65（54，74） 187/113 23.7（21.6，26.5） 117 ± 22 63 ± 15 80 ± 16 2.7 ± 2.1 26 ± 7 10 ± 4 54（15，130）
28 d死亡组 498 66（56，74） 307/191 23.2（20.4，25.6） 108 ± 20 60 ± 14 76 ± 14 3.4 ± 2.3 28 ± 8 13 ± 3 25（0，80）
χ2/t/Z值   0.925 0.014 2.633 6.011 2.927 4.343 4.143 4.676 10.110 5.973
P值   0.355 0.906 0.008 < 0.001 0.004 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001
组别 例数 CRRT病因[例（%）] CRRT剂量（mL/kg） 钾离子（mmol/L） 碳酸氢盐（mmol/L， ± s） 0 h磷酸根离子（mg/L）
少尿 代谢性酸中毒 容量超负荷 血尿 高钾血症 其他
28 d生存组 300 15（5.0） 53（17.7） 91（30.3） 63（21.0） 33（11.0） 45（15.0） 36.4（34.5，38.9） 4.50（4.10，5.20） 17 ± 5 155（117，196）
28 d死亡组 498 25（5.0） 131（26.3） 116（23.3） 121（24.3） 48（9.6） 57（11.4） 36.8（33.9，39.5） 4.50（4.00，5.25） 17 ± 6 177（130，228）
χ2/t/Z值   12.711 0.530 0.494 1.287 4.146
P值   0.026 0.596 0.621 0.199 < 0.001
组别 例数 24 h磷酸根离子（mg/L） 白细胞计数（× 106/L） 血红蛋白（g/L） 尿素氮（mg/L） 血肌酐（mg/L） 白蛋白（g/L， ± s） C反应蛋白（mg/L） 肾小球滤过率（mL/min） 尿素氮血肌酐比 C反应蛋白白蛋白比
28 d生存组 300 111（89，139） 12.19（8.87，19.29） 94（84，111） 470（340，700） 25.4（16.7，35.5） 27 ± 6 54（18，146） 23.2（15.6，38.5） 18.0（13.7，26.6） 18.6（6.6，54.2）
28 d死亡组 498 143（108，190） 10.69（4.79，18.30） 94（82，106） 530（360，750） 22.5（16.3，31.1） 25 ± 5 74（20，176） 28.3（18.7，39.2） 22.2（16.5，75.3） 29.4（8.0，75.3）
χ2/t/Z值   7.994 3.618 1.663 1.905 2.638 5.375 1.857 2.788 4.823 2.727
P值   < 0.001 < 0.001 0.096 0.057 0.008 < 0.001 0.063 0.005 < 0.001 0.006
图1 Lasso回归临床特征值筛选及模型构建与评估注：1 mmHg = 0.133 kPa；ROC.受试者工作特征；λ.正则化参数；SOFA.序贯器官衰竭估计；APACHE.急性病生理学和长期健康评价；a图为临床特征值的系数曲线；b图为10倍交叉验证选择最合适的临床特征值；c图为列线图模型；d图为不同危险因素模型ROC曲线
表2 影响脓毒症继发AKI患者28 d死亡的多因素logistic回归分析
变量 回归系数 标准误 Wald值 OR值 95%CI P值
白蛋白（g/L） -0.529 0.206 6.572 0.589 0.393 ~ 0.883 0.010
24 h磷酸根离子（mg/L） 0.341 0.070 23.526 1.406 1.225 ~ 1.613 < 0.001
SOFA评分（分） 0.210 0.035 36.207 1.234 1.152 ~ 1.321 < 0.001
血肌酐（mg/L） -0.257 0.082 9.797 0.773 0.658 ~ 0.908 0.002
平均动脉压（mmHg） -0.007 0.012 0.348 0.993 0.971 ~ 1.016 0.555
尿素氮肌酐比 0.017 0.008 4.368 1.017 1.001 ~ 1.034 0.037
收缩压（mmHg） -0.018 0.008 5.142 0.982 0.967 ~ 0.998 0.023
2 h尿量（mL） -0.003 0.001 6.400 0.997 0.995 ~ 0.999 0.011
常量 1.017 1.017 0.945 2.766 - 0.331
图2 危险因素模型的评估注：APACHE.急性病生理学和长期健康评价；a图为训练集列线图模型校准曲线；b图为验证集列线图模型校准曲线；c图为决策曲线；d图为Jung模型校准曲线；校准曲线实际曲线为模型预测曲线，偏倚矫正后曲线为Bootstrap验证后曲线
图3 基于列线图模型的不同危险因素组Kaplan-Meier曲线注：a图为28 d Kaplan-Meier曲线；b图为90 d Kaplan-Meier曲线；低风险组为死亡风险≤ 20%，中风险组为> 20% ~ < 80%，高风险组为≥ 80%
1
Uchino S, Kellum JA, Bellomo R, et al. Acute renal failure in critically ill patients: a multinational, multicenter study[J]. JAMA, 2005, 294 (7): 813-818.
2
Bagshaw SM, Uchino S, Bellomo R, et al. Septic acute kidney injury in critically ill patients: clinical characteristics and outcomes[J]. Clin J Am Soc Nephrol, 2007, 2 (3): 431-439.
3
Morgera S, Slowinski T, Melzer C, et al. Renal replacement therapy with high-cutoff hemofilters: impact of convection and diffusion on cytokine clearances and protein status[J]. Am J Kidney Dis, 2004, 43 (3): 444-453.
4
Lins RL, Elseviers MM, Van der Niepen P, et al. Intermittent versus continuous renal replacement therapy for acute kidney injury patients admitted to the intensive care unit: results of a randomized clinical trial[J]. Nephrol Dial Transplant, 2009, 24 (2): 512-518.
5
Miao H, Shi J, Wang C, et al. Continuous renal replacement therapy in pediatric severe sepsis: a propensity score-matched prospective multicenter cohort study in the PICU[J]. Crit Care Med, 2019, 47 (10): e806-e813.
6
Jung SY, Kwon J, Park S, et al. Phosphate is a potential biomarker of disease severity and predicts adverse outcomes in acute kidney injury patients undergoing continuous renal replacement therapy[J]. PLoS One, 2018, 13 (2): e0191290.
7
Mehta RL, Kellum JA, Shah SV, et al. Acute Kidney Injury Network: report of an initiative to improve outcomes in acute kidney injury[J]. Crit Care, 2007, 11 (2): R31.
8
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.
9
Mayeux PR, MacMillan-Crow LA. Pharmacological targets in the renal peritubular microenvironment: implications for therapy for sepsis-induced acute kidney injury[J]. Pharmacol Ther, 2012, 134 (2): 139-155.
10
Leedahl DD, Frazee EN, Schramm GE, et al. Derivation of urine output thresholds that identify a very high risk of AKI in patients with septic shock[J]. Clin J Am Soc Nephrol, 2014, 9 (7): 1168-1174.
11
Jaques DA, Spahr L, Berra G, et al. Biomarkers for acute kidney injury in decompensated cirrhosis: a prospective study[J]. Nephrology (Carlton), 2019, 24 (2): 170-180.
12
Priyanka P, Zarbock A, Izawa J, et al. The impact of acute kidney injury by serum creatinine or urine output criteria on major adverse kidney events in cardiac surgery patients[J]. J Thorac Cardiovasc Surg, 2021, 162 (1): 143-151.e7.
13
Burgess LG, Goyal N, Jones GM, et al. Evaluation of acute kidney injury and mortality after intensive blood pressure control in patients with intracerebral hemorrhage[J]. J Am Heart Assoc, 2018, 7 (8): e008439.
14
Fitzgerald JC, Ross ME, Thomas NJ, et al. Association of early hypotension in pediatric sepsis with development of new or persistent acute kidney injury[J]. Pediatr Nephrol, 2021, 36 (2): 451-461.
15
Walsh M, Devereaux PJ, Garg AX, et al. Relationship between intraoperative mean arterial pressure and clinical outcomes after noncardiac surgery: toward an empirical definition of hypotension[J]. Anesthesiology, 2013, 119 (3): 507-515.
16
Monk TG, Bronsert MR, Henderson WG, et al. Association between intraoperative hypotension and hypertension and 30-day postoperative mortality in noncardiac surgery [published correction appears in Anesthesiology, 2016, 124 (3): 741-742][J]. Anesthesiology, 2015, 123 (2): 307-319.
17
Sun LY, Wijeysundera DN, Tait GA, et al. Association of intraoperative hypotension with acute kidney injury after elective noncardiac surgery[J]. Anesthesiology, 2015, 123 (3): 515-523.
18
van Waes JA, van Klei WA, Wijeysundera DN, et al. Association between intraoperative hypotension and myocardial injury after vascular surgery[J]. Anesthesiology, 2016, 124 (1): 35-44.
19
Bae SJ, Lee SH, Yun SJ, et al. Comparison of IVC diameter ratio, BUN/creatinine ratio and BUN/albumin ratio for risk prediction in emergency department patients[J]. Am J Emerg Med, 2021 (47): 198-204.
20
Deng L, Wang C, Qiu S, et al. Association between blood urea nitrogen-to-creatinine ratio and three-month outcome in patients with acute ischemic stroke[J]. Curr Neurovasc Res, 2019, 16 (2): 166-172.
21
Wu KH, Shih HA, Hung MS, et al. The association between blood urea nitrogen to creatinine ratio and mortality in patients with upper gastrointestinal bleeding[J]. Arab J Gastroenterol, 2018, 19 (4): 143-147.
22
Qian H, Tang C, Yan G. Predictive value of blood urea nitrogen/creatinine ratio in the long-term prognosis of patients with acute myocardial infarction complicated with acute heart failure[J]. Medicine (Baltimore), 2019, 98 (11): e14845.
23
Uchino S, Bellomo R, Goldsmith D. The meaning of the blood urea nitrogen/creatinine ratio in acute kidney injury[J]. Clin Kidney J, 2012, 5 (2): 187-191.
24
陈敏华，朱峥，金君，等. B型利钠肽在预测重症患者急性肾损伤发生发展中的临床价值[J/CD].中华危重症医学杂志（电子版），2022，15（4）：279-284.
25
Patel A, Laffan MA, Waheed U, et al. Randomised trials of human albumin for adults with sepsis: systematic review and meta-analysis with trial sequential analysis of all-cause mortality [published correction appears in BMJ, 2014, (349): g4850][J]. BMJ, 2014 (349): g4561.
26
Padkins M, Breen T, Anavekar N, et al. Association between albumin level and mortality among cardiac intensive care unit patients[J]. J Intensive Care Med, 2021, 36 (12): 1475-1482.
27
Zhou H, Wang A, Meng X, et al. Low serum albumin levels predict poor outcome in patients with acute ischaemic stroke or transient ischaemic attack[J]. Stroke Vasc Neurol, 2021, 6 (3): 458-466.
28
Foley RN. Phosphate levels and cardiovascular disease in the general population[J]. Clin J Am Soc Nephrol, 2009, 4 (6): 1136-1139.
29
de Fornasari ML, Dos Santos Sens YA. Replacing phosphorus-containing food additives with foods without additives reduces phosphatemia in end-stage renal disease patients: a randomized clinical trial[J]. J Ren Nutr, 2017, 27 (2): 97-105.
[1] 韩圣瑾, 周正武, 翁云龙, 黄鑫. 碳酸氢钠林格液联合连续性肾脏替代疗法对创伤合并急性肾损伤患者炎症水平及肾功能的影响[J]. 中华危重症医学杂志(电子版), 2023, 16(05): 376-381.
[2] 张秋彬, 张楠, 林清婷, 徐军, 朱华栋, 姜辉. 急性胰腺炎合并急性肾损伤患者的预后评估[J]. 中华危重症医学杂志(电子版), 2023, 16(05): 382-389.
[3] 陈永庄, 莫小乔, 谢天. 心血管事件患者术后30 d死亡风险决策树模型的构建与评估——基于少数类样本合成过采样技术算法[J]. 中华危重症医学杂志(电子版), 2023, 16(05): 390-398.
[4] 张晓燕, 肖东琼, 高沪, 陈琳, 唐发娟, 李熙鸿. 转录因子12过表达对脓毒症相关性脑病大鼠大脑皮质的保护作用及其机制[J]. 中华妇幼临床医学杂志(电子版), 2023, 19(05): 540-549.
[5] 姚咏明. 如何精准评估烧伤脓毒症患者免疫状态[J]. 中华损伤与修复杂志(电子版), 2023, 18(06): 552-552.
[6] 张俊, 罗再, 段茗玉, 裘正军, 黄陈. 胃癌预后预测模型的研究进展[J]. 中华普通外科学文献(电子版), 2023, 17(06): 456-461.
[7] 唐旭, 韩冰, 刘威, 陈茹星. 结直肠癌根治术后隐匿性肝转移危险因素分析及预测模型构建[J]. 中华普外科手术学杂志(电子版), 2024, 18(01): 16-20.
[8] 王万川, 麦广智, 张晓槟, 冯家豪. 腹腔镜下LAR术治疗低位直肠癌保肛术中钉仓数量与术后LARS的关系[J]. 中华普外科手术学杂志(电子版), 2023, 17(05): 491-496.
[9] 贾成朋, 王代宏, 陈华, 孙备. 可切除性胰腺癌预后术前预测模型的建立及应用[J]. 中华普外科手术学杂志(电子版), 2023, 17(05): 566-570.
[10] 杨静, 顾红叶, 赵莹莹, 孙梦霞, 查园园, 王琪. 老年血液透析患者短期死亡的影响因素及列线图预测模型的预测作用[J]. 中华肾病研究电子杂志, 2023, 12(05): 254-259.
[11] 李青霖, 宋仁杰, 周飞虎. 一种重型劳力性热射病相关急性肾损伤小鼠模型的建立与探讨[J]. 中华肾病研究电子杂志, 2023, 12(05): 265-270.
[12] 任加发, 邬步云, 邢昌赢, 毛慧娟. 2022年急性肾损伤领域基础与临床研究进展[J]. 中华肾病研究电子杂志, 2023, 12(05): 276-281.
[13] 易成, 韦伟, 赵宇亮. 急性肾脏病的概念沿革[J]. 中华临床医师杂志(电子版), 2023, 17(08): 906-910.
[14] 谭睿, 王晶, 於江泉, 郑瑞强. 脓毒症中高密度脂蛋白、载脂蛋白A-I和血清淀粉样蛋白A的作用研究进展[J]. 中华临床医师杂志(电子版), 2023, 17(06): 749-753.
[15] 王亚丹, 吴静, 黄博洋, 王苗苗, 郭春梅, 宿慧, 王沧海, 王静, 丁鹏鹏, 刘红. 白光内镜下结直肠肿瘤性质预测模型的构建与验证[J]. 中华临床医师杂志(电子版), 2023, 17(06): 655-661.
阅读次数
全文


摘要

版权所有 中华医学会 中华医学电子音像出版社有限责任公司  京ICP备14006079号-1  网络出版服务许可证：(署)网出证(京)字第075号