基于LASSO-Cox回归构建列线图模型预测机械通气患者的压力性损伤风险

A nomogram model based on LASSO-Cox regression to predict pressure injury risk in mechanically ventilated patients

目的构建列线图模型预测重症监护室(intensive care unit,ICU)机械通气患者压力性损伤(pressure injuries,PI)的发生风险。方法回顾性收集2020年1月1日至2023年3月15日复旦大学附属中山医院ICU接受机械通气患者的数据作为训练集,2023年10月1日至2023年12月11日同一医院ICU接受机械通气患者的数据作为外部验证集。基于LASSO回归和Cox比例风险模型筛选PI的风险变量,构建列线图模型。绘制受试者工作特征(receiver operating characteristic,ROC)曲线并计算曲线下面积(area under the curve,AUC)以评价模型区分度,绘制校准曲线及决策曲线分析(decision curve analysis,DCA)评价模型校准度和临床适用性。将验证集数据代入列线图模型进行外部验证。结果训练集共纳入580例机械通气患者,84例(14.5%)发生PI。LASSO回归和Cox比例风险模型共筛选10个变量,构建列线图模型。ROC曲线显示,预测机械通气患者发生PI的AUC为0.830。校准曲线和DCA曲线提示模型校准度和预测效能良好。外部验证集共100例患者,12例发生PI,AUC为0.870,校准曲线和DCA曲线显示模型性能良好。结论基于LASSO-Cox回归构建的列线图模型预测性能较好,可用于机械通气患者PI高危人群的筛查。

Objective To construct a nomogram model to predict the risk of pressure injuries(PI) in mechanically ventilated patients in the intensive care unit(ICU). Methods Clinical data of mechanically ventilated patients in the ICU of Zhongshan Hospital, Fudan University from January 1, 2020 to March 15, 2023 were retrospectively collected as the training set, and data from ICU of the same hospital from October 1, 2023 to December 11, 2023 were collected as the external validation set. Risk variables for PI were selected using LASSO regression and Cox proportional hazards model, and a nomogram model was constructed. Receiver operating characteristic(ROC) curve was plotted, and the area under the curve(AUC) was calculated to evaluate the model. Calibration curve and decision curve analysis(DCA) were used to assess the model's calibration and clinical applicability. The external validation was performed using the validation set data. Results A total of 580 mechanically ventilated patients were included in the training set, with 84 cases(14.5%) of PI. LASSO regression and Cox proportional hazards model selected 10 variables to construct the nomogram model. The ROC curve showed an AUC of 0.830 for predicting PI in mechanically ventilated patients. Calibration curve and DCA indicated good calibration and predictive performance of the model. The external validation set included 100 patients, with 12 cases of PI, and the AUC was 0.870. Calibration curve and DCA showed good model performance.Conclusions The nomogram model based on LASSO-Cox regression has good predictive performance and can be used to screen high-risk population for PI in mechanically ventilated patients.