脓毒症相关急性肾损伤风险预测的研究现状及进展
Research Status and Advances in Risk Prediction of Sepsis-Associated Acute Kidney Injury
DOI: 10.12677/acm.2026.161128, PDF, HTML, XML,   
作者: 李俊岑:内蒙古科技大学包头医学院,内蒙古 包头;马文录*:内蒙古国药北方医院肾内科,内蒙古 包头
关键词: 急性肾损伤人工智能机器学习围手术期风险预测Acute Kidney Injury Artificial Intelligence Machine Learning Perioperative Period Risk Prediction
摘要: 急性肾损伤(AKI)是脓毒症患者最常见且最严重的器官功能障碍之一,具有发生率高、病死率高、医疗费用昂贵等特点。脓毒症相关AKI (Sepsis-associated AKI, SA-AKI)占重症监护病房(ICU)中AKI病例的40%~50%,是导致患者死亡的独立危险因素。传统的AKI预测方法存在准确性不足、时效性差等局限性,难以在肾功能恶化早期进行有效识别。近年来,机器学习算法在SA-AKI风险预测中的应用日益广泛,显示出良好的预测性能和临床应用前景。本文综述了SA-AKI的流行病学特征、发病机制、诊断标准以及机器学习算法在脓毒症相关AKI预测中的研究进展,分析了当前研究的不足与挑战,并展望了未来的发展方向。
Abstract: Acute kidney injury (AKI) is one of the most common and serious organ dysfunction in patients with sepsis, which has the characteristics of high incidence, high mortality and expensive medical expenses. Sepsis-associated AKI (SA-AKI) accounts for 40%~50% of AKI cases in intensive care unit (ICU), and it is an independent risk factor for patients’ death. The traditional AKI prediction method has some limitations, such as insufficient accuracy and poor timeliness, and it is difficult to effectively identify the early deterioration of renal function. In recent years, machine learning algorithm has been widely used in SA-AKI risk prediction, showing good prediction performance and clinical application prospect. This paper summarizes the epidemiological characteristics, pathogenesis, diagnostic criteria and research progress of machine learning algorithm in the prediction of Sepsis-associated AKI, analyzes the shortcomings and challenges of current research, and looks forward to the future development direction.
文章引用:李俊岑, 马文录. 脓毒症相关急性肾损伤风险预测的研究现状及进展[J]. 临床医学进展, 2026, 16(1): 972-983. https://doi.org/10.12677/acm.2026.161128

1. 引言

急性肾损伤(AKI)是指在数小时至数天内肾功能急剧下降的临床综合征,会导致各种毒素和代谢产物在体内蓄积,造成内环境紊乱,以血清肌酐升高、尿量减少为主要表现[1]。脓毒症是由感染引起的宿主反应失调所导致的危及生命的器官功能障碍,是重症监护病房(ICU)患者死亡的主要原因之一[2]。AKI是脓毒症最常见的器官功能障碍表现,据统计,约45%~70%的脓毒症患者会发生AKI,而在脓毒性休克患者中,这一比例可高达70%~80% [3]

脓毒症相关AKI (S-AKI)不仅显著增加患者的短期死亡率,还会导致住院时间延长、医疗费用增加,并可能进展为慢性肾脏病,严重影响患者的长期预后[4]。随着全球脓毒症发病率的上升和人口老龄化加剧,S-AKI已成为危重症医学领域的重大挑战。然而,S-AKI常以血清肌酐(serum creatinine, SCr)和尿量变化进行诊断,这些指标往往在肾损伤发生48~72小时后才出现明显变化,属于肾功能的“滞后指标”,错过了最佳的预防和治疗时机。

近年来,随着人工智能技术的快速发展,机器学习算法在S-AKI预测领域展现出巨大潜力。通过整合患者的人口学特征、生命体征、实验室检查、感染指标、器官功能评分等多维数据,机器学习模型能够识别复杂的非线性关系和交互作用,为实现S-AKI的早期识别和精准预测提供了新的途径,有望改善脓毒症患者的临床管理和预后。

2. 脓毒症相关AKI的流行病学与临床特征

2.1. SA-AKI的发生率和病死率

脓毒症相关AKI是ICU中最常见的AKI类型,占所有AKI病例的40%~50%。多项大规模流行病学研究显示,脓毒症患者中AKI的发生率在不同研究中存在差异,这主要与研究人群、诊断标准、地区医疗水平等因素有关。一项纳入170万住院患者的回顾性研究发现,脓毒症患者的AKI发生率为45.2%,而非脓毒症患者仅为15.8% [5]。在重症脓毒症和脓毒性休克患者中,AKI的发生率进一步上升,可达60%~70%。

SA-AKI的病死率显著高于其他原因所致的AKI。研究表明,发生AKI的脓毒症患者住院病死率为30%~50%,而未发生AKI的脓毒症患者病死率仅为10%~20% [6]。AKI的严重程度与病死率呈正相关,根据KDIGO分期,1期AKI患者的病死率约为20%~25%,2期为35%,3期可高达50%~60%。此外,需要肾脏替代治疗(renal replacement therapy, RRT)的SA-AKI患者病死率更高,可达60%。

值得注意的是,即使是轻度的AKI (KDIGO 1期)也与不良预后密切相关。一项Meta分析纳入了42项研究共283,751例患者,结果显示即使是SCr轻微升高(26.5 μmol/L)也与住院病死率增加67%相关[7]。这提示早期识别和干预轻度AKI具有重要临床意义。

2.2. 高危人群特征

某些特定人群发生SA-AKI的风险更高。年龄是重要的危险因素之一,≥65岁的老年患者由于肾脏储备功能下降、合并症增多,SA-AKI的发生率显著高于年轻患者。既往存在慢性肾脏病(chronic kidney disease, CKD)的患者是SA-AKI的高危人群,其发生率是肾功能正常者的2~3倍[8]

合并症如糖尿病、高血压、心力衰竭、肝硬化等也是SA-AKI的独立危险因素。糖尿病患者由于存在糖尿病肾病、血管病变等基础疾病,更易发生SA-AKI,且预后更差。免疫抑制状态(如器官移植术后、恶性肿瘤化疗期间)的患者发生脓毒症时,AKI风险也明显升高。

在感染来源方面,腹腔感染、肺部感染是导致SA-AKI最常见的原因。革兰阴性菌感染,特别是产生内毒素的细菌,更容易引起严重的全身炎症反应和AKI [9]。此外,感染性休克、需要血管活性药物支持、机械通气等因素均与SA-AKI的发生密切相关。

2.3. 疾病负担与医疗费用

SA-AKI不仅增加患者的近期死亡风险,还对长期预后产生深远影响。幸存者中有相当比例会发展为慢性肾脏病甚至终末期肾病(end-stage renal disease, ESRD),需要长期透析治疗。研究显示,SA-AKI患者出院后1年内发展为CKD的风险是未发生AKI者的2~3倍,进展至ESRD的风险增加5~10倍[10]。这种“AKI-CKD连续谱”的概念近年来受到广泛关注。

从经济负担角度看,SA-AKI显著增加了医疗费用。AKI患者的住院时间更长,ICU停留时间延长,需要更多的医疗资源和支持治疗。一项美国的研究显示,发生AKI的脓毒症患者平均住院费用较未发生AKI者增加40%~60%,需要RRT的患者费用更高。在中国,SA-AKI患者的平均住院费用也显著高于非AKI患者,约增加2~3万元人民币[11]

3. SA-AKI的预测因子与传统风险模型

3.1. 临床危险因素

一些专注于识别AKI重要危险因素的研究已确定,患者易感性和暴露因素对于AKI的发展同样重要。患者的易感性包括年龄、性别、种族和合并症。在所有合并症中,CKD是AKI的主要危险因素,因其与自主调节功能失调、肾脏储备丧失以及对肾毒性药物的敏感性有关。糖尿病、高血压、心血管疾病、高尿酸血症、肥胖和肝病也被认为是AKI的危险因素[12]。脓毒症、肾毒性药物、外科干预和休克已被确定为AKI的促成因素。一项多中心国际横断面AKI-EPI研究报告,脓毒症、血容量不足和肾毒性药物暴露是危重病患者AKI最常见的三种原因[13]。在一些暴露中,身体状况较差的患者发生AKI的概率可能会更高,例如老年患者心脏手术SA-AKI。但AKI的风险会因为身体状况和肾毒性暴露的不同而不同,这使得准确地评估风险变得困难。

3.2. 生物标志物(包括传统与新型)

目前临床上用于AKI预测的方法主要包括临床评分系统、生物标志物检测和传统统计学预测模型。常用的临床评分系统包括SOFA评分、SAPS II评分等,这些评分系统简单易用,但预测准确性有限,在AKI预测中的AUC值通常在0.6~0.7之间[14]。Bell等(2015) [15]在一项包含15,872例患者的前瞻性队列研究中发现,入院时白蛋白 < 30 g/L的患者AKI发生率为28.3%,显著高于白蛋白 ≥ 35 g/L患者的10.1% (P < 0.001)。多因素分析显示,低白蛋白血症是AKI的独立危险因素(OR = 2.84, 95%CI: 2.31-3.49)。白蛋白通过以下机制影响肾功能:(1) 维持血浆胶体渗透压,当白蛋白 < 25 g/L时,有效循环血量显著减少,肾灌注压下降;(2) 抗氧化作用,白蛋白含有丰富的巯基,能够清除自由基,减少氧化应激损伤;(3) 载体功能,影响药物的分布和清除,低白蛋白状态下游离药物浓度增加,可能加重肾毒性。

Caraceni等研究证实,白蛋白具有重要的抗炎作用。在炎症状态下,白蛋白不仅合成减少,其分子结构也发生改变,抗氧化和载体功能下降,形成恶性循环[16]。除白蛋白外,其他营养指标也与AKI风险相关。Zhang等[17]研究了前白蛋白与AKI的关系,发现前白蛋白 < 150 mg/L的患者AKI发生率显著高于正常组,前白蛋白半衰期短(2~3天),能够更敏感地反映蛋白质合成状态。白介素-6 (IL-6)作为重要的促炎细胞因子,在AKI发病机制中发挥关键作用。Skrypnyk等[18]研究阐明了IL-6参与AKI的具体机制:(1) 激活内皮细胞,增加血管通透性,导致毛细血管渗漏和间质水肿;(2) 促进血小板聚集和血栓形成,加重肾脏微循环障碍;(3) 直接激活肾小管上皮细胞的凋亡程序,通过caspase-3途径诱导细胞死亡。降钙素原(PCT)在脓毒症相关AKI预测中具有重要价值。Gao等[19]研究发现PCT > 2.0 ng/mL的患者AKI发生率显著高于PCT < 0.5 ng/mL患者,PCT不仅反映感染严重程度,还与肾脏微循环障碍密切相关。超敏C反应蛋白(hs-CRP)作为急性期反应蛋白,其预测价值在多项研究中得到证实。Liu等(2009)的荟萃分析显示,CRP每增加10 mg/L,AKI风险增加15% (95%CI: 8%~23%) [20]。炎症反应的级联效应Gómez等(2014)提出了炎症介导AKI的“级联反应”理论:初始的炎症刺激(如感染、缺血)激活单核-巨噬细胞系统,释放TNF-α、IL-1β、IL-6等促炎因子;这些因子进一步激活血管内皮细胞和肾小管上皮细胞,产生粘附分子和趋化因子;最终导致中性粒细胞浸润、血管通透性增加、微血栓形成,引起肾脏损伤[21]

由于血清肌酐水平和尿量都是非特异性和不准确的AKI标志物,多种新型生物标志物已被研究用于及时预测或诊断AKI。以下新型生物标志物已被确定用于AKI的早期检测:胱抑素C、中性粒细胞明胶酶相关脂质运载蛋白、肾损伤分子1、肝型脂肪酸结合蛋白、尿血管紧张素原(AGT)和钙蛋白。中性粒细胞明胶酶相关脂质运载蛋白(NGAL)是近年来研究最多的AKI早期生物标志物[22] [23]。Bennett等(2008)在心脏手术患者中的研究发现,术后2小时血清NGAL>150μg/L能够预测48小时内AKI的发生,AUC为0.88,显著优于血肌酐(AUC = 0.52)。NGAL的优势在于其在肾小管损伤后2~6小时即可升高,具有良好的早期预警价值[24]。肾损伤分子-1 (KIM-1)是肾小管损伤的特异性标志物。Han等(2008)研究了201例重症患者,发现尿KIM-1 > 2.5 ng/mL的患者AKI发生率为68.3%,AUC为0.83。KIM-1的特点是在肾小管损伤时特异性表达,在健康人群中几乎检测不到[25]。胱抑素C (Cys C)不受年龄、性别、肌肉量影响,是评估肾功能的理想指标。在AKI预测方面,Zhang等(2012)发现胱抑素C > 1.2 mg/L的患者AKI发生率为45.6%,显著高于正常组的12.3%。胱抑素C的优势在于其在肾功能轻度下降时即可升高,具有更好的早期诊断价值[26]

尽管近期研究确定的新型AKI生物标志物大大改善并使AKI的早期检测成为可能,但是在临床上使用这些生物标志物仍然有很多困难。Vanmassenhove等使用新型的血清和尿液生物标志物早期诊断AKI仍较为麻烦,尤其是在AKI原因未明确时[27]。另一个困难是,新型生物标志物的检测可能并不普遍,或较为昂贵。Marx等认为,临床很难简单地通过一个通用血清或尿液生物标志物来确定AKI的风险、诊断、严重程度和预后,以及区分AKI的原因并监测其进展[28]。AKI是非均质性、复杂的疾病,其具有广泛的原因和病理生理学,为了对AKI进行标准化诊断,需要几种能够涵盖AKI不同方面的生物标志物或生物标志物组合似乎是合理的[29] [30]。检查多种新型生物标志物或使用生物标志物组合来评估患者的状况可能会导致早期准确预测或诊断AKI的成本更高。

3.3. 基于上述因素的传统风险评分模型

AKI的危险因素确定以后,研究者们就着手去运用独立的AKI预测因子组合,考量各自的影响,再实施外部检验,从而创建风险评分。准确的风险预测评分必须能够辨识出风险较大的病患,并指引医生去预防,诊治疾病,针对不同患者群体来评价AKI风险,不同的评分系统就此创建起来,这些预测模型包括年龄、性别、基线肾功能以及合并症状况,在考量手术种类、药物、程序时,可以加入特定的预测因素。

Mehran风险评分于2004年被提出,它被用来分析经皮冠状动脉介入治疗之后患者的AKI风险以及肾脏替代治疗需求,按照2016年所展开的后续外部验证,此系统在预测接受冠状动脉造影的急性冠状动脉综合征患者对比剂引发的肾病时表现尚可[31] [32]。大规模队列研究显示手术是AKI的主要原因,AKI发生率从创伤手术的25%到心脏或主动脉手术的50%不等[33]。此外,心脏手术在所有类型的手术中AKI发生率最高,范围从2%到50%,透析依赖率为1%到6% [34] [35];因此,为计划接受心脏手术的患者建立几个AKI风险识别预测模型并不令人惊讶。最早的评分系统EuroSCORE基于1999年发表的欧洲多中心数据,2010年的年龄、肌酐和射血分数价值(ACEF)评分也基于欧洲数据库的数据[36]。短期风险(胸外科医师协会,STS)评分于2008年使用美国胸外科医师协会国家数据库的数据创建;该评分用于评估成人术前心脏手术风险,专业人员已保留并修改了该预测模型[37] [38]。在一项外部验证研究中,196名患者接受了二尖瓣修复术,比较了他们的STS和ACEF评分;STS肾衰竭评分在预测2期和3期AKI方面最准确。此外,该研究发现ACEF评分在所有AKI预测中表现出与STS肾衰竭评分相似的AUROC(ACEF和STS评分AUROC分别为0.758和0.797),但ACEF评分仅包括三个预测因素:年龄、肌酐和射血分数;因此,ACEF评分对临床医生更方便。在另一项比较单独冠状动脉旁路移植手术AKI术前风险模型的研究中,EuroSCORE II、STS评分和ACEF评分在预测3期AKI方面都表现适当;此外,ACEF评分在预测SA-AKI方面表现出令人满意的判别力,AUROC为0.781 [39]

除了合并症和急性疾病状况外,根据先前的研究,种族和流行病学因素也显示出对AKI发生率的影响。Mathioudakis等基于美国国家数据库报告,黑人与白人相比,年龄和性别调整后的AKI几率高50% (几率比:1.51;95% CI 1.37~1.66),在对多种AKI相关危险因素进行额外调整后,黑人种族与AKI风险增加之间的关联仍然存在[40]。2013年,一项专注于全球AKI发生率的meta分析报告,根据KDIGO标准的AKI合并率在世界各地显示出差异。根据世界地理区域和国家经济模式以及纬度,AKI合并率在南美洲与北美洲相比更高(29.6%对24.5%),南欧与北欧相比更高(31.5%对14.7%),南亚与西亚或东亚相比更高(23.7%对16.7%对14.7%)。位于赤道以南与以北国家的AKI合并率更高(27.0%对22.6%),此外,该研究还显示在医疗总支出占GDP >10%与≤5%的国家中,AKI发生率较高(25.2%对14.5%) [41]

考虑到种族和流行病学对AKI发生率的影响,一些研究人员已经根据其国家健康保险研究数据库验证了他们的评分,以实现高预测性能。一个例子是ADVANCIS评分,用于预测因冠状动脉疾病接受经皮冠状动脉介入治疗(PCI)患者的AKI。ADVANCIS评分使用八个临床参数(年龄、糖尿病、呼吸机使用、既往AKI、干预血管数量、CKD、IABP使用和心源性休克),评分范围从0到22;此外,ADVANCIS评分 ≥ 6与住院死亡率风险增加相关[42]。除了对流行病学因素进行风险预测模型的修正之外,研究人员还将新的生物标志物纳入一些现代AKI预测评分系统中,作为预测因素并评估了生物标志物与患者临床信息之间的相关性。Zhou等人将尿NGAL和尿AGT作为危险因素,建立急性失代偿性心衰患者AKI预测评分[43]

尽管已经创建了各种评分系统来处理不同类型的临床情况,但是大多数预测模型都只能是单点AKI预测模型,比如预测某种类型的手术之后或者在使用造影剂之前发生的AKI几率,无法实时显示变化,而且,一些评分系统包含了很多因素,如基础状况、临床数据和新的生物标志物,这使得它们在临床上使用起来过于复杂,由于信息技术的进步,一些医院已经在他们的医疗信息系统(MIS)中集成了这些预测系统,这些临床风险评估工具由于能够自动分析数据而越来越受欢迎,由于种族、基因、疾病发病率和药物在各个国家之间的差异,使用MIS和风险预测评分的组合可能会使使用当地数据库的数据来评估AKI风险和肾脏替代治疗的需求成为可能。

4. 基于机器学习算法建立AKI预测模型

随着MIS越来越普遍,提供自动化电子警报系统(e-alerts)变得越来越可行,基于数据系统,使用算法分析患者的电子记录和临床数据,以预测是否存在早期或亚临床的AKI [44]。Park等开发了一个有自动肾科医生咨询的AKI警报系统,当患者的血清肌酐水平从基线升高到至少1.5倍或0.3 mg/dL时,临床医生可以向肾科医生发出自动肾科医生咨询。该研究报告,在引入e-alert系统后,肾脏科医生的早期诊断准确率明显提升(调整OR,6.13;95% CI,4.80~7.82),严重AKI事件显著降低(调整OR,0.75;95% CI,0.64~0.89);但死亡率没有受到影响(调整HR,1.07;95% CI,0.68~1.68) [45]。另一项研究在ICU患者中使用了e-alert系统,当e-alert系统筛选血清肌酐数据并根据KDIGO标准定义检测到可能的AKI事件时,临床医生收到“弹出”消息。尽管在该研究中,AKI e-alert系统的敏感性、特异性、约登指数和准确性分别为99.8%、97.7%、97.5%和98.1%,e-alert组中AKI诊断患病率和肾脏科诊断患病率高于非e-alert组。两组在透析、肾功能恢复或死亡的患病率方面没有显著差异[46]。2017年,一项系统综述得出结论,e-alerts系统既不降低死亡率(比值比[OR],1.05;95% CI,0.84~1.31)也不降低透析治疗的发生率(OR,1.20;95% CI,0.91~1.57) [47] [48]。该meta分析中包含的所有六项研究都仅使用血清肌酐变化作为e-alerts的触发因素,如前段所述,血清肌酐变化既不是肾损伤的敏感性也不是特异性标志物。除了血清肌酐作为AKI标志物的局限性外,e-alerts系统在用于没有基线肾功能的患者和那些基线肌酐水平较高、肌酐水平小幅变化后肾功能变化更显著的CKD患者时面临挑战;临床医生在收到e-alerts后为患者提供各种各样的进一步护理。为了在收到e-alerts后提供标准化和循证的临床护理,建立了护理包。最新的指南没有为AKI提供特定的管理选择,治疗策略主要是支持性的。在危重病患者中,遵循KDIGO指南详细说明的液体管理、避免肾毒性药物、监测血清肌酐水平和血流动力学以及转诊专家后,AKI的发生和严重程度得到降低。当e-alert系统与护理包结合,分析患者病史,检测患者尿液样本,建立AKI临床诊断,制定治疗和检测方案,并寻求肾脏科医生建议时,几项研究报告了医院获得性AKI和AKI相关死亡率以及住院天数的减少[49] [50]。机器学习算法需求量大且需要大量数据。借助大型EMR数据库和强大的计算硬件,学者们扩展了机器学习的应用。最近,AI也与各种机器学习算法一起应用,特别是深度神经网络。

机器学习算法凭借其强大的数据处理能力和模式识别能力,在AKI预测领域展现出显著优势。根据算法原理和复杂程度,可将其分为传统机器学习算法和深度学习算法两大类[51] [52]

传统机器学习算法中,随机森林表现最为突出。在一项涉及671例心脏手术患者的研究中,随机森林模型在SA-AKI预测中获得了0.839的AUC值,集成模型(RF + XGBoost)的AUC进一步提升至0.843 [53]。XGBoost算法在多项研究中也显示出优异性能,在非体外循环冠状动脉旁路移植术后AKI预测中AUC达到0.87,在脓毒症患者AKI预测中AUC为0.817,显著优于传统评分系统[54] [55]。支持向量机和决策树等算法虽然性能相对较低,但具有较好的可解释性,在临床应用中仍有一定价值。

深度学习算法在AKI预测中的应用日益广泛,特别是循环神经网络(RNN)在处理时间序列数据方面具有独特优势。一项基于15,564例患者的研究显示,RNN模型在心胸外科术后AKI预测中获得了0.893的AUC值,在与经验丰富的临床医师的直接对比中,RNN显著优于医师判断(AUC 0.901 vs 0.745, p < 0.001) [56]。卷积神经网络(CNN)在ICU患者48小时AKI预测中也表现良好,AUC达到0.86 [57]。DeepAKI等深度可解释网络的开发,进一步提高了深度学习模型的临床实用性[58]

动态预测是机器学习算法的重要优势之一。与传统的静态预测不同,动态机器学习模型能够每小时更新AKI风险评估,对未来48小时内任何程度AKI的预测AUC为0.82,对2级或更严重AKI的预测AUC达0.95,在89%的病例中能在临床检测前预测AKI发生[59]。这种实时风险评估能力为临床早期干预提供了宝贵的时间窗口。

在国外研究方面,美国和欧洲在机器学习AKI预测领域起步较早,研究基础相对雄厚。Google DeepMind团队开发的AKI预测模型在美国退伍军人事务部的数据中表现优异,但缺乏独立的外部验证[60]。荷兰开发的AKIpredictor在前瞻性临床验证中表现良好,与医师预测性能相当但校准度更佳[61]

2016年,Thottakkara团队采用支持向量机算法预测脓毒症AKI,预测的AUC值提升至0.744 [62]。2017年,Cheng等人采用随机森林(Random Forest, RF)进行AKI预测,AUC值为0.765 [63]。2018年,Lee团队采用梯度提升算法预测心脏手术后AKI,AUC值明显提升,达0.78 [64]。同年,Mohamadlou等[65]使用RF模型预测AKI,取得了优异的AUC值(0.84),算法持续改进。特别值得关注的是,Koyner团队在2018年通过梯度增强机算法预测2期肾损伤,在24h内预测AKI的AUC值为0.90,48小时内AUC值为0.87,创下了该领域的最佳预测效果[66]。然而,部分研究使用私有数据集,这在一定程度上限制了模型的普适性和可重复性,成为未来研究需要突破的重要方向。

近年来,机器学习方法在ICU患者AKI预测领域取得了显著进展。Chiofolo等(2019) [67]率先采用RF算法构建了连续的AKI风险评分模型,其ROC达到0.88,但该研究仅限于ICU患者群体。Zhang等(2019) [68]则利用XGBoost模型对少尿AKI患者的容量反应性进行预测,取得了AUROC值表现优异(0.860)。张渊等(2019) [69]采用LightGBM算法预测ICU患者AKI风险,模型的准确率高达0.89。池锐彬等(2020) [70]基于生物标志物特征,运用决策树算法预测重症AKI,模型精准度达到86.0%。He等(2019) [71]通过比较5种不同机器学习方法构建预测模型,结果显示各模型的AUROC值介于0.720至0.764之间。值得注意的是,深度学习方法在该领域展现出独特优势,其无监督学习能力以及高效处理大规模数据的特点,使其在处理复杂数据时显著优于传统回归分析方法。综上所述,多种机器学习方法在ICU患者AKI预测中均展现出较高的准确性,其中深度学习方法在处理大规模复杂数据时具有突出优势。在国内研究方面,中国学者在机器学习AKI预测领域也取得了重要进展[72] [73]。广西医科大学团队开发的GBDT模型在非体外循环冠状动脉旁路移植术后AKI预测中表现优异[74]。然而,国内研究在多中心验证、前瞻性研究和国际合作方面仍有待加强。

从研究趋势看,机器学习在AKI预测领域的应用呈现几个特点:首先是从单一算法向集成学习发展,通过融合多种算法优势提高预测性能;其次是从静态预测向动态监测发展,实现实时风险评估;第三是从黑盒模型向可解释模型发展,SHAP分析等可解释性技术的应用日益广泛;第四是从单中心研究向多中心验证发展,提高模型的泛化能力[75]

5. 小结和未来研究进展

机器学习算法在SA-AKI风险预测中展现出显著优势,特别是在预测准确性、动态监测能力和大数据处理方面。深度学习模型表现最为优异,能够显著优于传统预测方法和临床医师判断。然而,当前研究仍面临外部验证、标准化、可解释性等挑战。机器学习在SA-AKI预测领域的发展前景广阔,未来研究应重点关注以下几个方向。首先是建立标准化的验证框架,包括统一的AKI定义、一致的性能评估指标和标准化的报告规范,以提高研究的可重现性和可比性。其次是加强多中心合作研究,开展大规模、多国家的前瞻性验证研究,提高模型的泛化能力和临床实用性。在技术发展方面,联邦学习等新兴技术有望解决数据隐私和多中心合作的问题,实现在保护患者隐私的前提下进行大规模模型训练。可解释性技术的进一步发展将提高模型的临床接受度,而边缘计算技术的应用将使实时预测成为可能。临床转化是机器学习AKI预测研究的最终目标。未来需要开发集成化的临床决策支持系统,将AKI预测模型嵌入电子健康记录系统,实现无缝的临床工作流程整合。同时,需要建立相应的质量控制和监管机制,确保AI辅助诊断的安全性和有效性。

NOTES

*通讯作者。

参考文献

[1] 徐丽斌. 电子预警系统对医院获得性急性肾损伤早期诊断的临床价值[Z]. 内蒙古自治区人民医院, 2021-09-08.
[2] Kuo, G., Yang, S.Y., Chuang, S.S., Fan, P.C., et al. (2016) Using Acute Kidney Injury Severity and Scoring Systems to Predict Outcome in Patients with Burn Injury. Journal of the Formosan Medical Association, 115, 1046-1052. [Google Scholar] [CrossRef] [PubMed]
[3] Zimmerman, L.P., Reyfman, P.A., Smith, A.D.R., Zeng, Z., Kho, A., Sanchez-Pinto, L.N., et al. (2019) Early Prediction of Acute Kidney Injury Following ICU Admission Using a Multivariate Panel of Physiological Measurements. BMC Medical Informatics and Decision Making, 19, Article No. 16. [Google Scholar] [CrossRef] [PubMed]
[4] Khera, R., Haimovich, J., Hurley, N.C., McNamara, R., Spertus, J.A., Desai, N., et al. (2021) Use of Machine Learning Models to Predict Death after Acute Myocardial Infarction. JAMA Cardiology, 6, 633-641. [Google Scholar] [CrossRef] [PubMed]
[5] Wang, H.E., Muntner, P., Chertow, G.M. and Warnock, D.G. (2012) Acute Kidney Injury and Mortality in Hospitalized Patients. American Journal of Nephrology, 35, 349-355. [Google Scholar] [CrossRef] [PubMed]
[6] Bagshaw, S.M., Uchino, S., Bellomo, R., et al. (2007) Septic Acute Kidney Injury in Critically Ill Patients: Clinical Characteristics and Outcomes. Clinical Journal of the American Society of Nephrology, 2, 431-439. [Google Scholar] [CrossRef] [PubMed]
[7] Coca, S.G., Yusuf, B., Shlipak, M.G., Garg, A.X. and Parikh, C.R. (2009) Long-Term Risk of Mortality and Other Adverse Outcomes after Acute Kidney Injury: A Systematic Review and Meta-Analysis. American Journal of Kidney Diseases, 53, 961-973. [Google Scholar] [CrossRef] [PubMed]
[8] Bellomo, R., Kellum, J.A., Ronco, C., Wald, R., Martensson, J., Maiden, M., et al. (2017) Acute Kidney Injury in Sepsis. Intensive Care Medicine, 43, 816-828. [Google Scholar] [CrossRef] [PubMed]
[9] Lopes, J.A. and Jorge, S. (2013) The RIFLE and AKIN Classifications for Acute Kidney Injury: A Critical and Comprehensive Review. Clinical Kidney Journal, 6, 8-14. [Google Scholar] [CrossRef] [PubMed]
[10] Wald, R., McArthur, E., Adhikari, N.K.J., Bagshaw, S.M., Burns, K.E.A., Garg, A.X., et al. (2015) Changing Incidence and Outcomes Following Dialysis-Requiring Acute Kidney Injury among Critically Ill Adults: A Population-Based Cohort Study. American Journal of Kidney Diseases, 65, 870-877. [Google Scholar] [CrossRef] [PubMed]
[11] Chronopoulos, A., Rosner, M.H., Cruz, D.N. and Ronco, C. (2010) Acute Kidney Injury in Elderly Intensive Care Patients: A Review. Intensive Care Medicine, 36, 1454-1464. [Google Scholar] [CrossRef] [PubMed]
[12] Hahn, K., Kanbay, M., Lanaspa, M.A., Johnson, R.J. and Ejaz, A.A. (2017) Serum Uric Acid and Acute Kidney Injury: A Mini Review. Journal of Advanced Research, 8, 529-536. [Google Scholar] [CrossRef] [PubMed]
[13] Sgura, F.A., Bertelli, L., Monopoli, D., Leuzzi, C., Guerri, E., Spartà, I., et al. (2010) Mehran Contrast-Induced Nephropathy Risk Score Predicts Short-and Long-Term Clinical Outcomes in Patients with St-Elevation-Myocardial Infarction. Circulation: Cardiovascular Interventions, 3, 491-498. [Google Scholar] [CrossRef] [PubMed]
[14] Martinez, D.A., Levin, S.R., Klein, E.Y., Parikh, C.R., Menez, S., Taylor, R.A., et al. (2020) Early Prediction of Acute Kidney Injury in the Emergency Department with Machine-Learning Methods Applied to Electronic Health Record Data. Annals of Emergency Medicine, 76, 501-514. [Google Scholar] [CrossRef] [PubMed]
[15] Bell, S., Dekker, F.W., Vadiveloo, T., Marwick, C., Deshmukh, H., Donnan, P.T., et al. (2015) Risk of Postoperative Acute Kidney Injury in Patients Undergoing Orthopaedic Surgery—Development and Validation of a Risk Score and Effect of Acute Kidney Injury on Survival: Observational Cohort Study. British Medical Journal, 351, h5639. [Google Scholar] [CrossRef] [PubMed]
[16] Caraceni, P., Tufoni, M. and Bonavita, M.E. (2013) Clinical Use of Albumin. Blood Transfusion, 11, S18.
[17] Zhang, L., Xue, S., Wu, M. and Dong, D. (2022) Performance of Urinary Liver-Type Fatty Acid-Binding Protein in Diabetic Nephropathy: A Meta-Analysis. Frontiers in Medicine, 9, Article 914587. [Google Scholar] [CrossRef] [PubMed]
[18] Skrypnyk, N.I., Gist, K.M., Okamura, K., Montford, J.R., You, Z., Yang, H., et al. (2020) Il-6-Mediated Hepatocyte Production Is the Primary Source of Plasma and Urine Neutrophil Gelatinase-Associated Lipocalin during Acute Kidney Injury. Kidney International, 97, 966-979. [Google Scholar] [CrossRef] [PubMed]
[19] Gao, L.I., Zhong, X., Jin, J., Li, J. and Meng, X. (2020) Potential Targeted Therapy and Diagnosis Based on Novel Insight into Growth Factors, Receptors, and Downstream Effectors in Acute Kidney Injury and Acute Kidney Injury-Chronic Kidney Disease Progression. Signal Transduction and Targeted Therapy, 5, Article No. 9. [Google Scholar] [CrossRef] [PubMed]
[20] Liu, K.D., Altmann, C., Smits, G., Krawczeski, C.D., Edelstein, C.L., Devarajan, P., et al. (2009) Serum Interleukin-6 and Interleukin-8 Are Early Biomarkers of Acute Kidney Injury and Predict Prolonged Mechanical Ventilation in Children Undergoing Cardiac Surgery: A Case-Control Study. Critical Care, 13, R104. [Google Scholar] [CrossRef] [PubMed]
[21] Gomez, H., Ince, C., de Backer, D., et al. (2014) A Unified Theory of Sepsis-Induced Acute Kidney Injury: Inflammation, Microcirculatory Dysfunction, Bioenergetics, and the Tubular Cell Adaptation to Injury. Shock, 41, 3-11.
[22] Grams, M.E., Sang, Y., Coresh, J., Ballew, S., Matsushita, K., Molnar, M.Z., et al. (2016) Acute Kidney Injury after Major Surgery: A Retrospective Analysis of Veterans Health Administration Data. American Journal of Kidney Diseases, 67, 872-880. [Google Scholar] [CrossRef] [PubMed]
[23] Dudoignon, E., Dépret, F. and Legrand, M. (2019) Is the Renin-Angiotensin-Aldosterone System Good for the Kidney in Acute Settings? Nephron, 143, 179-183. [Google Scholar] [CrossRef] [PubMed]
[24] Bennett, M., Dent, C.L., Ma, Q., Dastrala, S., Grenier, F., Workman, R., et al. (2008) Urine NGAL Predicts Severity of Acute Kidney Injury after Cardiac Surgery: A Prospective Study. Clinical Journal of the American Society of Nephrology, 3, 665-673. [Google Scholar] [CrossRef] [PubMed]
[25] Han, W.K., Wagener, G., Zhu, Y., Wang, S. and Lee, H.T. (2009) Urinary Biomarkers in the Early Detection of Acute Kidney Injury after Cardiac Surgery. Clinical Journal of the American Society of Nephrology, 4, 873-882. [Google Scholar] [CrossRef] [PubMed]
[26] Zhang, Z., Lu, B., Sheng, X. and Jin, N. (2011) Cystatin C in Prediction of Acute Kidney Injury: A Systemic Review and Meta-Analysis. American Journal of Kidney Diseases, 58, 356-365. [Google Scholar] [CrossRef] [PubMed]
[27] Vanmassenhove, J., Vanholder, R., Nagler, E. and Van Biesen, W. (2013) Urinary and Serum Biomarkers for the Diagnosis of Acute Kidney Injury: An In-Depth Review of the Literature. Nephrology Dialysis Transplantation, 28, 254-273. [Google Scholar] [CrossRef] [PubMed]
[28] Marx, D., Metzger, J., Pejchinovski, M., Gil, R.B., Frantzi, M., Latosinska, A., et al. (2018) Proteomics and Metabolomics for AKI Diagnosis. Seminars in Nephrology, 38, 63-87. [Google Scholar] [CrossRef] [PubMed]
[29] Kashani, K., Cheungpasitporn, W. and Ronco, C. (2017) Biomarkers of Acute Kidney Injury: The Pathway from Discovery to Clinical Adoption. Clinical Chemistry and Laboratory Medicine (CCLM), 55, 1074-1089. [Google Scholar] [CrossRef] [PubMed]
[30] Lameire, N.H., Bagga, A., Cruz, D., De Maeseneer, J., Endre, Z., Kellum, J.A., et al. (2013) Acute Kidney Injury: An Increasing Global Concern. The Lancet, 382, 170-179. [Google Scholar] [CrossRef] [PubMed]
[31] Abellás-Sequeiros, R.A., Raposeiras-Roubín, S., Abu-Assi, E., González-Salvado, V., Iglesias-Álvarez, D., Redondo-Diéguez, A., et al. (2016) Mehran Contrast Nephropathy Risk Score: Is It Still Useful 10 Years Later? Journal of Cardiology, 67, 262-267. [Google Scholar] [CrossRef] [PubMed]
[32] Uchino, S., Kellum, J.A., Bellomo, R., Doig, G.S., et al. (2005) Acute Renal Failure in Critically Ill Patients: A Multinational, Multicenter Study. Journal of the American Medical Association, 294, 813-818. [Google Scholar] [CrossRef] [PubMed]
[33] Chang, C.H., Lee, C.C., Chen, S.W., Fan, P.C., et al. (2016) Predicting Acute Kidney Injury Following Mitral Valve Repair. International Journal of Medical Sciences, 13, 19-24. [Google Scholar] [CrossRef] [PubMed]
[34] Wang, Y. and Bellomo, R. (2017) Cardiac Surgery-Associated Acute Kidney Injury: Risk Factors, Pathophysiology and Treatment. Nature Reviews Nephrology, 13, 697-711. [Google Scholar] [CrossRef] [PubMed]
[35] Nashef, S.A.M., Roques, F., Michel, P., Gauducheau, E., Lemeshow, S. and Salamon, R. (1999) European System for Cardiac Operative Risk Evaluation (Euroscore). European Journal of Cardio-Thoracic Surgery, 16, 9-13. [Google Scholar] [CrossRef] [PubMed]
[36] Wykrzykowska, J.J., Garg, S., Onuma, Y., de Vries, T., Goedhart, D., Morel, M., et al. (2011) Value of Age, Creatinine, and Ejection Fraction (ACEF Score) in Assessing Risk in Patients Undergoing Percutaneous Coronary Interventions in the ‘All-Comers’ LEADERS Trial. Circulation: Cardiovascular Interventions, 4, 47-56. [Google Scholar] [CrossRef] [PubMed]
[37] Shahian, D.M., Jacobs, J.P., Badhwar, V., Kurlansky, P.A., Furnary, A.P., Cleveland, J.C., et al. (2018) The Society of Thoracic Surgeons 2018 Adult Cardiac Surgery Risk Models: Part 1—Background, Design Considerations, and Model Development. The Annals of Thoracic Surgery, 105, 1411-1418. [Google Scholar] [CrossRef] [PubMed]
[38] O’Brien, S.M., Feng, L., He, X., Xian, Y., Jacobs, J.P., Badhwar, V., et al. (2018) The Society of Thoracic Surgeons 2018 Adult Cardiac Surgery Risk Models: Part 2—Statistical Methods and Results. The Annals of Thoracic Surgery, 105, 1419-1428. [Google Scholar] [CrossRef] [PubMed]
[39] Wendt, D., Thielmann, M., Kahlert, P., Kastner, S., Price, V., Al-Rashid, F., et al. (2014) Comparison between Different Risk Scoring Algorithms on Isolated Conventional or Transcatheter Aortic Valve Replacement. The Annals of Thoracic Surgery, 97, 796-802. [Google Scholar] [CrossRef] [PubMed]
[40] Mathioudakis, N.N., Giles, M., Yeh, H.C., et al. (2016) Racial Differences in Acute Kidney Injury of Hospitalized Adults with Diabetes. Journal of Diabetes and its Complications, 30, 1129-1136. [Google Scholar] [CrossRef] [PubMed]
[41] Susantitaphong, P., Cruz, D.N., Cerda, J., Abulfaraj, M., et al. (2013) Acute Kidney Injury Advisory Group of the American Society of Nephrology. World Incidence of AKI: A Meta-Analysis. Clinical Journal of the American Society of Nephrology, 8, 1482-1493. [Google Scholar] [CrossRef] [PubMed]
[42] Fan, P.C., Chen, T.H., Lee, C.C., et al. (2018) ADVANCIS Score Predicts Acute Kidney Injury after Percutaneous Coronary Intervention for Acute Coronary Syndrome. International Journal of Medical Sciences, 15, 528-535. [Google Scholar] [CrossRef] [PubMed]
[43] Zhou, L.Z., Yang, X.B., Guan, Y., Xu, X., Tan, M.T., Hou, F.F., et al. (2016) Development and Validation of a Risk Score for Prediction of Acute Kidney Injury in Patients with Acute Decompensated Heart Failure: A Prospective Cohort Study in China. Journal of the American Heart Association, 5, e004035. [Google Scholar] [CrossRef] [PubMed]
[44] Cheungpasitporn, W. and Kashani, K. (2016) Electronic Data Systems and Acute Kidney Injury. In: Contributions to Nephrology, S. Karger AG, 73-83. [Google Scholar] [CrossRef] [PubMed]
[45] Park, S., Baek, S.H., Ahn, S., Lee, K.H., et al. (2018) Impact of Electronic Acute Kidney Injury (AKI) Alerts with Automated Nephrologist Consultation on Detection and Severity of AKI: A Quality Improvement Study. American Journal of Kidney Diseases, 71, 9-19. [Google Scholar] [CrossRef] [PubMed]
[46] Wu, Y., Chen, Y., Li, S., Dong, W., Liang, H., Deng, M., et al. (2018) Value of Electronic Alerts for Acute Kidney Injury in High-Risk Wards: A Pilot Randomized Controlled Trial. International Urology and Nephrology, 50, 1483-1488. [Google Scholar] [CrossRef] [PubMed]
[47] Lachance, P., Villeneuve, P., Rewa, O.G., Wilson, F.P., Selby, N.M., Featherstone, R.M., et al. (2017) Association between E-Alert Implementation for Detection of Acute Kidney Injury and Outcomes: A Systematic Review. Nephrology Dialysis Transplantation, 32, 265-272. [Google Scholar] [CrossRef] [PubMed]
[48] Lachance, P., Villeneuve, P., Wilson, F.P., Selby, N.M., Featherstone, R., Rewa, O., et al. (2016) Impact of E-Alert for Detection of Acute Kidney Injury on Processes of Care and Outcomes: Protocol for a Systematic Review and Meta-Analysis. BMJ Open, 6, e011152. [Google Scholar] [CrossRef] [PubMed]
[49] Kolhe, N.V., Reilly, T., Leung, J., Fluck, R.J., Swinscoe, K.E., Selby, N.M., et al. (2016) A Simple Care Bundle for Use in Acute Kidney Injury: A Propensity Score-Matched Cohort Study. Nephrology Dialysis Transplantation, 31, 1846-1854. [Google Scholar] [CrossRef] [PubMed]
[50] Kolhe, N.V., Staples, D., Reilly, T., Merrison, D., Mcintyre, C.W., Fluck, R.J., et al. (2015) Impact of Compliance with a Care Bundle on Acute Kidney Injury Outcomes: A Prospective Observational Study. PLOS ONE, 10, e0132279. [Google Scholar] [CrossRef] [PubMed]
[51] Hodgson, L.E., Roderick, P.J., Venn, R.M., Yao, G.L., Dimitrov, B.D. and Forni, L.G. (2018) The ICE-AKI Study: Impact Analysis of a Clinical Prediction Rule and Electronic AKI Alert in General Medical Patients. PLOS ONE, 13, e0200584. [Google Scholar] [CrossRef] [PubMed]
[52] Kate, R.J., Perez, R.M., Mazumdar, D., Pasupathy, K.S. and Nilakantan, V. (2016) Prediction and Detection Models for Acute Kidney Injury in Hospitalized Older Adults. BMC Medical Informatics and Decision Making, 16, Article No. 39. [Google Scholar] [CrossRef] [PubMed]
[53] Tseng, P.Y., Chen, Y.T., Wang, C.H., et al. (2020) Prediction of the Development of Acute Kidney Injury Following Cardiac Surgery by Machine Learning. Critical Care, 24, Article No. 478. [Google Scholar] [CrossRef] [PubMed]
[54] Zheng, S., Li, Y., Luo, C., Chen, F., Ling, G. and Zheng, B. (2023) Machine Learning for Predicting the Development of Postoperative Acute Kidney Injury after Coronary Artery Bypass Grafting without Extracorporeal Circulation. Cardiovascular Innovations and Applications, 7, 1-16. [Google Scholar] [CrossRef
[55] Zhou, H., Liu, L., Zhao, Q., Jin, X., Peng, Z., Wang, W., et al. (2023) Machine Learning for the Prediction of All-Cause Mortality in Patients with Sepsis-Associated Acute Kidney Injury during Hospitalization. Frontiers in Immunology, 14, Article 1140755. [Google Scholar] [CrossRef] [PubMed]
[56] Rank, N., Pfahringer, B., Kempfert, J., Stamm, C., Kühne, T., Schoenrath, F., et al. (2020) Deep-Learning-Based Real-Time Prediction of Acute Kidney Injury Outperforms Human Predictive Performance. npj Digital Medicine, 3, Article No. 139. [Google Scholar] [CrossRef] [PubMed]
[57] Le, S., Allen, A., Calvert, J., Palevsky, P.M., Braden, G., Patel, S., et al. (2021) Convolutional Neural Network Model for Intensive Care Unit Acute Kidney Injury Prediction. Kidney International Reports, 6, 1289-1298. [Google Scholar] [CrossRef] [PubMed]
[58] Yang, M., Liu, S., Hao, T., Ma, C., Chen, H., Li, Y., et al. (2024) Development and Validation of a Deep Interpretable Network for Continuous Acute Kidney Injury Prediction in Critically Ill Patients. Artificial Intelligence in Medicine, 149, Article 102785. [Google Scholar] [CrossRef] [PubMed]
[59] Alfieri, F., Ancona, A., Tripepi, G., Rubeis, A., Arjoldi, N., Finazzi, S., et al. (2023) Continuous and Early Prediction of Future Moderate and Severe Acute Kidney Injury in Critically Ill Patients: Development and Multi-Centric, Multi-National External Validation of a Machine-Learning Model. PLOS ONE, 18, e0287398. [Google Scholar] [CrossRef] [PubMed]
[60] Tomašev, N., Glorot, X., Rae, J.W., Zielinski, M., Askham, H., Saraiva, A., et al. (2019) A Clinically Applicable Approach to Continuous Prediction of Future Acute Kidney Injury. Nature, 572, 116-119. [Google Scholar] [CrossRef] [PubMed]
[61] Flechet, M., Falini, S., Bonetti, C., Güiza, F., Schetz, M., Van den Berghe, G., et al. (2019) Machine Learning versus Physicians’ Prediction of Acute Kidney Injury in Critically Ill Adults: A Prospective Evaluation of the AKIpredictor. Critical Care, 23, Article No. 282. [Google Scholar] [CrossRef] [PubMed]
[62] Thottakkara, P., Ozrazgat-Baslanti, T., Hupf, B.B., Rashidi, P., Pardalos, P., Momcilovic, P., et al. (2016) Application of Machine Learning Techniques to High-Dimensional Clinical Data to Forecast Postoperative Complications. PLOS ONE, 11, e0155705. [Google Scholar] [CrossRef] [PubMed]
[63] Cheng, P., Waitman, L.R., Hu, Y., et al. (2017) Predicting Inpatient Acute Kidney Injury over Different Time Horizons: How Early and Accurate. Annual Symposium Proceedings, Washington, 4-8 November 2017, 565-574.
[64] Lee, H.C., Yoon, H.K., Nam, K., et al. (2018) Derivation and Validation of Machine Learning Approaches to Predict Acute Kidney Injury after Cardiac Surgery. Journal of Clinical Medicine, 7, Article 322. [Google Scholar] [CrossRef] [PubMed]
[65] Mohamadlou, H., Lynn-Palevsky, A., Barton, C., Chettipally, U., Shieh, L., Calvert, J., et al. (2018) Prediction of Acute Kidney Injury with a Machine Learning Algorithm Using Electronic Health Record Data. Canadian Journal of Kidney Health and Disease, 5, Article 2054358118776326. [Google Scholar] [CrossRef] [PubMed]
[66] Koyner, J.L., Carey, K.A., Edelson, D.P. and Churpek, M.M. (2018) The Development of a Machine Learning Inpatient Acute Kidney Injury Prediction Model. Critical Care Medicine, 46, 1070-1077. [Google Scholar] [CrossRef] [PubMed]
[67] Chiofolo, C., Chbat, N., Ghosh, E., Eshelman, L. and Kashani, K. (2019) Automated Continuous Acute Kidney Injury Prediction and Surveillance: A Random Forest Model. Mayo Clinic Proceedings, 94, 783-792. [Google Scholar] [CrossRef] [PubMed]
[68] Zhang, Z.H., Ho, K.M. and Hong, Y.C. (2019) Machine Learning for the Prediction of Volume Responsiveness in Patients with Oliguric Acute Kidney Injury in Critical Care. Critical Care, 23, 1-10. [Google Scholar] [CrossRef] [PubMed]
[69] 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.
[70] 池锐彬, 梁美华, 邹启明, 等. 基于生物标志物预测重症患者急性肾损伤决策树模型的构建和验证研究[J]. 中华危重病急救医学, 2020, 32(6): 721-725.
[71] He, J.Q., Hu, Y., Zhang, X.Z., et al. (2019) Multi-Perspective Predictive Modeling for Acute Kidney Injury in General Hospital Populations Using Electronic Medical Records. JAMIA Open, 2, 115-122. [Google Scholar] [CrossRef] [PubMed]
[72] Li, Y.K., Yao, L., Mao, C.S., et al. (2018) Early Prediction of Acute Kidney Injury in Critical Care Setting Using Clinical Notes. 2018 IEEE International Conference on Bioinformatics and Biomedicine, Madrid, 3-6 December 2018, 683-686. [Google Scholar] [CrossRef] [PubMed]
[73] 朱道谋, 钟丽花, 陈彩华. 血清和尿NGAL、KIM-1、CysC对晚发型败血症新生儿急性肾损伤的早期预警价值[J]. 东南大学学报(医学版), 2021, 40(2): 176-182.
[74] 王林国. 急性创伤性颅脑损伤后急性肺损伤高危因素分析及早期预警指标研究[Z]. 桐庐县第一人民医院, 2021-03-12.
[75] 胡慧宇, 张敏, 周兴梅, 等. 尿微量白蛋白/尿肌酐比值预测体外循环心脏手术后急性肾损伤及预后的价值分析[J]. 中国现代医学杂志, 2020, 30(16): 33-38.