作者：赵文婷1，周大文1，杨晓梅1，郑红艳2

单位：

1.223002江苏省淮安市第二人民医院呼吸与危重症医学科　2.223002江苏省淮安市第二人民医院神经外科

单位（英文）：

1.Department of Respiratory and Critical Care Medicine, Huai'an Second People's Hospital, Huaian 223002, China2.Department of Neurosurgery, Huai'an Second People's Hospital, Huaian 223002, China

关键词：

机械通气；脱机失败；影响因素分析；列线图；预测

关键词（英文）：

Mechanical ventilations; Offline failure; Root cause analysis; Nomograms; Forecasting

中图分类号：

R 605.973

DOI：

10.12114/j.issn.1008-5971.2023.00.141

基金项目：

摘要：

目的　构建并验证重症机械通气患者脱机失败的风险预测列线图模型。方法　采用便利抽样法，选取2020年5月至2022年5月在淮安市第二人民医院重症医学科进行机械通气的患者670例为研究对象。按照7∶3的比例将患者分为建模组（n=469）及验证组（n=201）。根据脱机结果将建模组进一步分为失败亚组（n=88）和成功亚组（n=381）。自行设计基线资料调查表并统计患者基线资料，采用多因素Logistic回归分析探讨建模组重症机械通气患者脱机失败的影响因素；基于多因素Logistic回归分析结果，采用R软件构建重症机械通气患者脱机失败的风险预测列线图模型；采用Hosmer-Lemeshow检验及校准曲线评估该列线图模型的校准度，采用ROC曲线评估该列线图模型的区分度。结果　失败亚组机械通气时间≥7 d、入院24 h内最低急性生理学与慢性健康状况评分系统Ⅱ（APACHE Ⅱ）评分≥15分、入院24 h内最低脓毒症相关性器官功能衰竭评价（SOFA）评分≥6分者占比及通气后动脉血二氧化碳分压（PaCO2）、呼吸机相关性膈肌功能障碍（VIDD）发生率高于成功亚组，脱机前24 h内血清白蛋白低于成功亚组（P＜0.05）。多因素Logistic回归分析结果显示，机械通气时间、入院24 h内最低APACHE Ⅱ评分、入院24 h内最低SOFA评分、通气后PaCO2、VIDD是建模组重症机械通气患者脱机失败的影响因素（P＜0.05）。基于多因素Logistic回归分析结果，构建重症机械通气患者脱机失败的风险预测列线图模型。Hosmer-Lemeshow检验及校准曲线分析结果显示，该列线图模型预测建模组、验证组重症机械通气患者脱机失败的发生率分别与本组重症机械通气患者脱机失败的实际发生率比较，差异无统计学意义（χ2=7.650，P=0.468；χ2=7.465，P=0.487）。ROC曲线分析结果显示，该列线图模型预测建模组、验证组重症机械通气患者脱机失败的曲线下面积分别为0.870〔95%CI（0.836，0.903）〕、0.867〔95%CI（0.812，0.911）〕，灵敏度分别为74.47%、75.31%，特异度分别为87.19%、85.83%。结论　机械通气时间≥7 d、入院24 h内最低APACHE Ⅱ评分≥15分、入院24 h内最低SOFA评分≥6分、通气后PaCO2升高、VIDD是重症机械通气患者脱机失败的危险因素，基于以上危险因素构建的重症机械通气患者脱机失败的风险预测列线图模型具有良好的校准度、区分度，其对重症机械通气患者脱机失败风险具有良好的预测能力。

英文摘要：

Objective To construct and validate a nomogram model for predicting the risk of offline failure in patientswith severe mechanical ventilation. Methods A total of 670 patients with mechanical ventilation admitted to the Department ofIntensive Care Medicine of Huai'an Second People's Hospital from May 2020 to May 2022 were selected as the research objects byconvenience sampling method. Patients were divided into modeling group (n=469) and validation group (n=201) in a ratio of 7∶3.According to the offline results, the modeling group was further divided into failure subgroup (n=88) and success subgroup (n=381) .The baseline data questionnaire was designed and the baseline data of patients were collected. Multivariate Logistic regressionanalysis was used to explore the influencing factors of offline failure in patients with severe mechanical ventilation in the modelinggroup. Based on the results of multivariate Logistic regression analysis, R software was used to establish the nomogram model forpredicting the risk of offline failure in patients with severe mechanical ventilation. Hosmer-Lemeshow test and calibration curvewere used to evaluate the calibration degree of the nomogram model. ROC curve was used to evaluate the differentiation of thenomogram model. Results The proportion of mechanical ventilation duration ≥ 7 d, the lowest acute physiology and chronic health evaluation Ⅱ (APACHE Ⅱ) score ≥ 15 points within 24 h of admission, the lowest sepsis-related organ failure assessment(SOFA) score ≥ 6 points within 24 h of admission, partial pressure of arterial carbon dioxide (PaCO2) after ventilation and theincidence of ventilator-induced diaphragmatic dysfunction (VIDD) in the failure subgroup were higher than those in the successfulsubgroup, and the serum albumin within 24 h before offline was lower than that in the successful subgroup (P < 0.05) . MultivariateLogistic regression analysis showed that mechanical ventilation duration, the lowest APACHE Ⅱ score within 24 h of admission,the lowest SOFA score within 24 h of admission, PaCO2 after ventilation and VIDD were the influencing factors of offline failurein patients with severe mechanical ventilation in the modeling group (P < 0.05) . Based on the results of multivariate Logisticregression analysis, the nomogram model for predicting the risk of offline failure in patients with severe mechanical ventilation wasconstructed. Hosmer-Lemeshow test and calibration curve analysis results showed that there was no significant difference in theincidence of offline failure predicted by nomogram model in the modeling group and the verification group and the actual incidenceof offline failure in patients with severe mechanical ventilation, respectively (χ2=7.650, P=0.468; χ2=7.465, P=0.487) . ROCcurve analysis results showed that the area under the curve of nomogram model in predicting offline failure in patients with severemechanical ventilation in the modeling group and the verification group was 0.870 [95%CI (0.836, 0.903) ] and 0.867 [95%CI(0.812, 0.911) ] , the sensitivity was 74.47% and 75.31%, the specificity was 87.19% and 85.83%, respectively. ConclusionMechanical ventilation duration ≥ 7 d, the lowest APACHEⅡ score ≥ 15 points within 24 h of admission, the lowest SOFA score ≥6 points within 24 h of admission, increased PaCO2 after ventilation, VIDD are risk factors for offline failure in patients with severemechanical ventilation. The nomogram model for predicting the risk of offline failure in patients with severe mechanical ventilationconstructed based on the above risk factors has good calibration and differentiation, and has good prediction ability for the risk ofoffline failure in patients with severe mechanical ventilation.

参考文献：