文献速递重症综合管理与镇静镇痛消化、代谢与肾脏急症

连续肾脏替代治疗非计划中断风险预测模型的构建与外部验证

[Construction and external validation of a risk prediction model for unplanned interruption during continuous renal replacement therapy].
PubMed 文献 · Zhonghua Wei Zhong Bing Ji Jiu Yi Xue · 2024-01-01
一句话结论基于慢性病、血净导管功能、血流速度和血泵停转次数构建的CRRT非计划中断风险预测模型,内部验证AUC 0.952、外部验证AUC 0.919,可快速实时识别高危患者。

该研究回顾性纳入浙江医院ICU 2021年1月至2022年8月接受CRRT的331例危重成人患者,按是否发生非计划下机分组,比较潜在影响因素,经二元Logistic回归筛选独立影响因素并构建风险预测模型,用Hosmer-Lemeshow检验验证拟合优度、ROC曲线评价预测效度。模型嵌入医院ICU多功能电子病历系统后,前瞻性纳入2022年11月至2023年10月110例CRRT患者进行外部验证。结果显示慢性病、血液净化导管功能、血流速度、血泵停转次数为独立影响因素,内部验证AUC 0.952(95%CI 0.930-0.973,P=0.0030),外部验证AUC 0.919(95%CI 0.870-0.969,P=0.0043),模型预测效率较高,可为预防性护理提供参考。

要点

  • 回顾性建模队列331例CRRT危重患者中,非计划中断组93例、计划中断组238例;外部验证队列110例中,非计划中断组47例、计划中断组63例。
  • 二元Logistic回归显示慢性病(OR=3.063,95%CI 1.200-7.819)、血液净化导管功能(OR=4.429,95%CI 1.270-15.451)、血流速度(OR=0.928,95%CI 0.900-0.957)、血泵停转次数(OR=1.339,95%CI 1.231-1.457)为CRRT非计划中断的独立影响因素(均P<0.05)。
  • 内部验证模型AUC为0.952(95%CI 0.930-0.973,P=0.0030),敏感度88.2%、特异度89.9%,约登指数最大值为1.781。
  • 外部验证模型AUC为0.919(95%CI 0.870-0.969,P=0.0043),敏感度91.5%、特异度79.4%,约登指数最大值为1.709。
  • 单因素比较中,非计划中断组男性比例更低(80.6% vs. 91.6%)、慢性病比例更高(60.2% vs. 41.6%)、导管功能不良比例更高(31.2% vs. 6.3%)、CRRT启动前血小板计数更高[×10⁹/L:137(101,187)vs. 109(74,160)]、血流速度更低[mL/min:120(120,150)vs. 150(140,180)]、使用前稀释比例更高(37.6%

适用人群与场景

  • 浙江医院ICU接受CRRT治疗的危重成人患者,回顾性建模队列为2021年1月至2022年8月收治者,前瞻性外部验证队列为2022年11月至2023年10月收治者。
  • 按是否发生非计划下机分为计划中断组与非计划中断组。

诊断标准与关键阈值

  • 以二元Logistic回归筛选出的4个独立影响因素构建非计划中断风险预测模型。

研究设计与关键结果

  • 研究类型为回顾性研究构建模型并前瞻性外部验证;建模队列331例(计划中断组238例、非计划中断组93例),外部验证队列110例(计划中断组63例、非计划中断组47例)。
  • 单因素比较:非计划中断组男性比例更低(80.6% vs. 91.6%)、慢性病比例更高(60.2% vs. 41.6%)、血液净化导管功能不良比例更高(31.2% vs. 6.3%)、CRRT启动前血小板计数更高[×10⁹/L:137(101,187)vs. 109(74,160)]、血流速度更低[mL/min:120(120,150)vs. 150(140,180)]、使用前稀释比例更高(37.6% vs. 23.5%)、滤过分数更高[23.0%(17.5%,32.9%)vs. 19.1%(15.7%,22.6%)]、血泵停转次数更多[次:19(14,21)vs. 9(6,13)],差异均有统计学意义(均P<0.05)。
  • 二元Logistic回归:慢性病OR=3.063(95%CI 1.200-7.819)、血液净化导管功能OR=4.429(95%CI 1.270-15.451)、血流速度OR=0.928(95%CI 0.900-0.957)、血泵停转次数OR=1.339(95%CI 1.231-1.457)为独立影响因素(均P<0.05)。
  • 内部验证ROC曲线AUC=0.952(95%CI 0.930-0.973,P=0.0030),敏感度88.2%、特异度89.9%,约登指数最大值1.781。
  • 外部验证ROC曲线AUC=0.919(95%CI 0.870-0.969,P=0.0043),敏感度91.5%、特异度79.4%,约登指数最大值1.709。
  • 模型嵌入医院ICU多功能电子病历系统用于临床验证。

证据类型、等级与局限

  • 证据类型:回顾性研究构建模型并前瞻性外部验证。
  • 模型拟合优度经Hosmer-Lemeshow检验验证,预测效度经ROC曲线评价。
  • 局限:原文未报告局限性。

对床旁实践的改变

  • 作者认为该风险预测模型预测效率较高,可快速、实时识别CRRT非计划中断高危患者,为预防性护理提供参考。

提炼说明

  • 本条依据论文摘要整理;摘要未涉及「处置流程与药物剂量」「红旗征与不能漏的情况」「指南或共识出处」,故未列出这些栏目;全文请点「阅读原文」。
英文原文(摘要)· 点击展开

OBJECTIVE: To identify the independent factors of unplanned interruption during continuous renal replacement therapy (CRRT) and construct a risk prediction model, and to verify the clinical application effectiveness of the model.

METHODS: A retrospective study was conducted on critically ill adult patients who received CRRT treatment in the intensive care unit (ICU) of Zhejiang Hospital from January 2021 to August 2022 for model construction. According to whether unplanned weaning occurred, the patients were divided into two groups. The potential influencing factors of unplanned CRRT weaning in the two groups were compared. The independent influencing factors of unplanned CRRT weaning were screened by binary Logistic regression and a risk prediction model was constructed. The goodness of fit of the model was verified by a Hosmer-Lemeshow test and its predictive validity was evaluated by receiver operator characteristic curve (ROC curve). Then embed the risk prediction model into the hospital's ICU multifunctional electronic medical record system for severe illness, critically ill patients with CRRT admitted to the ICU of Zhejiang Hospital from November 2022 to October 2023 were prospectively analyzed to verify the model's clinical application effect.

RESULTS: (1) Model construction and internal validation: a total of 331 critically ill patients with CRRT were included to be retrospectively analyzed. Among them, there were 238 patients in planned interruption group and 93 patients in unplanned interruption group. Compared with the planned interruption group, the unplanned interruption group was shown as a lower proportion of males (80.6% vs. 91.6%) and a higher proportion of chronic diseases (60.2% vs. 41.6%), poor blood purification catheter function (31.2% vs. 6.3%), as a higher platelet count (PLT) before CRRT initiation [×109/L: 137 (101, 187) vs. 109 (74, 160)], lower level of blood flow rate [mL/min: 120 (120, 150) vs. 150 (140, 180)], higher proportion of using pre-dilution (37.6% vs. 23.5%), higher filtration fraction [23.0% (17.5%, 32.9%) vs. 19.1% (15.7%, 22.6%)], and frequency of blood pump stops [times: 19 (14, 21) vs. 9 (6, 13)], the differences of the above 8 factors between the two groups were statistically significant (all P < 0.05). Binary Logistic regression analysis showed that chronic diseases [odds ratio (OR) = 3.063, 95% confidence interval (95%CI) was 1.200-7.819], blood purification catheter function (OR = 4.429, 95%CI was 1.270-15.451), blood flow rate (OR = 0.928, 95%CI was 0.900-0.957), and frequency of blood pump stops (OR = 1.339, 95%CI was 1.231-1.457) were the independent factors for the unplanned interruption of CRRT (all P < 0.05). These 4 factors were used to construct a risk prediction model, and ROC curve analysis showed that the area under the curve (AUC) predicted by the model was 0.952 (95%CI was 0.930-0.973, P = 0.003 0), with a sensitivity of 88.2%, a specificity of 89.9%, and a maximum value of 1.781 for the Youden index. (2) External validation: prospective inclusion of 110 patients, including 63 planned interruption group and 47 unplanned interruption group. ROC curve analysis showed that the AUC of the risk prediction model was 0.919 (95%CI was 0.870-0.969, P = 0.004 3), with a sensitivity of 91.5%, a specificity of 79.4%, and a maximum value of the Youden index of 1.709.

CONCLUSIONS: The risk prediction model for unplanned interruption during CRRT has a high predictive efficiency, allowing for rapid and real-time identification of the high risk patients, thus providing references for preventative nursing.

来源:Zhonghua Wei Zhong Bing Ji Jiu Yi Xue,查看原文。版权归原作者所有。

参考来源

  1. Xu H, Ren Q, Zhu L, et al. [Construction and external validation of a risk prediction model for unplanned interruption during continuous renal replacement therapy]. Zhonghua Wei Zhong Bing Ji Jiu Yi Xue. 2024. PMID: 38845500. doi:10.3760/cma.j.cn121430-20231204-01045

供执业医师抢救与诊疗参考,不替代临床判断;剂量请结合体重、肝肾功能和本院药品规格复核,以指南原文为准。