ICU机械通气患者拔管后 48h再插管的风险预测模型构建与验证 [中文引用][英文引用]

目的 构建重症监护室（intensive care unit，ICU）机械通气患者拔管后 48h 再插管的风险预测模型，并进行内部验证。方法 选取 2018 年 1 月至 2024 年 6 月 1 815 例本市某三级甲等综合医院 ICU 机械通气患者为研究对象，采用计算机生成随机数的方法，按 8：2 的比例分为建模集（n=1 452）和验证集（n=363），并根据建模集机械通气拔管后 48h 是否需要再插管分为撤机成功组（n=1 142）和撤机失败组（n=310）。采用 LASSO 回归和多因素 Logistic 回归分析筛选建模集患者机械通气拔管后 48h 再插管的影响因素，并基于建模集影响因素构建列线图预测模型；采用简单数据集拆分法（hold-out）进行内部验证；采用受试者工作特征（receiver operating characteristic，ROC）、曲线下面积（area under the curve，AUC）、校准曲线、决策曲线评价建模集和验证集模型性能与临床适用性。结果 多因素结果显示，急性生理和慢性健康 -Ⅱ（acute physiology and chronic healthevaluation，APACHE-Ⅱ）评分高、合并心力衰竭、合并慢性阻塞性肺病（chronic obstructive pulmonary disease，COPD）、机械通气时间长、氧合指数低、最大吸气压力（maximum inspiratory pressure，MIP）小、白蛋白（albumin，ALB）低是患者拔管后 48h 再插管的风险因素（均 P<0.05）；建模集和验证集预测机械通气患者拔管后 48h 再插管的 AUC 分别为 0.868（95%CI：0.848~0.888）、0.831（95%CI：0.775~0.887），敏感度分别为 0.797、0.808，特异度分别为 0.783、0.766；校准曲线显示建模集和验证集 Brier 评分分别为 0.117 和 0.097；决策曲线结果表明，阈值在 0.10~0.92 内时模型的临床净获益高。结论 本研究构建的机械通气患者拔管后 48h 再插管的列线图预测模型具有较好的预测价值，能为临床决策提供参考依据。

Objective A risk prediction model was developed and verified to evaluate the prediction effect on re-intubation within 48 hours after extubation in the patients under mechanical ventilation in intensive care unit（ICU）. Methods A total of 1, 815 patients under mechanical ventilation in the ICU of a Tier-IIIA hospital from January 2018 to June 2024 were selected. The patients were randomly divided into a modelling set（n=1, 452）and a validation set（n=363）according to computer-generated random numbers, with a ratio of 8：2. Then according to the re-intubation within 48 hours after extubation in the modelling set, the patients were further divided into a weaning success group（n=1, 142）and a weaning failure group（n=310）. Lasso regression analysis and multivariate Logistic regression analysis were used to screen the independent factors influencing the re-intubation within 48 hours after extubation from mechanical ventilation. A nomogram prediction model was developed for the prediction, based on the identified independent influencing factors. Internal validation of the prediction model was performed using the simple hold-out method. The receiver operating characteristic （ROC）curve, calibration curve and decision curve were used to verify the predictive effect, calibration and clinical applicability of the model. Results The results of multiple factors showed that a high acute physiology and chronic health（APACHE-II）score, heart failure and chronic obstructive pulmonary disease（COPD）, long mechanical ventilation time, low oxygenation index, small maximum inspiratory pressure（MIP）and low albumin（ALB）were the independent factors influencing the re-intubation within 48 hours after extubation（all P<0.05）. The areas under the curve for predicting re-intubation within 48 hours after extubation in the modelling set and the validation set were 0.868（95%CI：0.848-0.888）and 0.831（95%CI：0.775-0.887）, respectively, with 79.70% and 80.80% of sensitivity and 78.30% and 76.60% of specification, respectively. The calibration curve showed that the Brier scores of the training set and the validation set were 0.117 and 0.097 respectively. The results of the decision curve indicated that the clinical net benefit of the model was high when the threshold probability was 0.10 to 0.92. Conclusion Based on independent risk factors, the nomogram prediction model for re-intubation within 48 hours after extubation in the patients under mechanical ventilation gives a good prediction and a reference for clinical decision-making.