重症俯卧位通气患者并发压力性损伤列线图模型的构建

Construction of nomogram model for predicting pressure injury in patients with severe prone position ventilation

目的构建重症俯卧位通气(PPV)患者并发压力性损伤(PI)的列线图模型。方法选取2019年8月至2022年12月在温州医科大学附属第一医院ICU行PPV治疗的200例重症患者作为训练集数据,根据PPV治疗期间是否发生PI分为PI组77例和非PI组123例。利用单因素及LASSO回归分析筛选影响因素并纳入多因素logistic回归分析,构建列线图模型。运用Calibration校准曲线、一致性指数(C-index)及Hosmer-Lemeshow检验评估预测效果。最后,纳入2023年2至6月的60例数据验证模型的准确性。结果多因素logistic回归分析显示,Braden评分、白蛋白、有饮酒史、机械通气时长、血管活性药物使用时长是均重症PPV患者并发PI的独立影响因素(均P<0.05)。构建模型的训练集和验证集C-index分别为0.763(95%CI:0.695~0.831)和0.732(95%CI:0.596~0.868);Calibration校准曲线证实模型有较好一致性,Hosmer-Lemeshow拟合优度检验的χ2值分别为10.801和8.697,均P>0.05。结论构建的重症PPV患者PI的列线图模型预测效果良好,使用便捷,可为临床提供参考。

Objective To develop a prediction nomogram for predicting pressure injury(PI)in critically ill patients undergoing prone position ventilation(PPV).Methods A total of 200 patients who underwent PPV at the First Affiliated Hospital of Wenzhou Medical University from August 2019 to December 2022 were selected for training set and 60 patients receiving PPV from February to June 2023 were selected as validation set.In training set PI occurred during in 77 cases(PI group)and did not occur in 123 cases(non-PI group).The influencing factors of PI were analyzed with univariate analysis,LASSO regression and multivariate logistic regression analysis.Based on the independent influencing factors,a nomogram for predicting PI in PPV patients was then constructed.The predictive performance was evaluated using Calibration curves,concordance index(C-index),and Hosmer-Lemeshow test.The accuracy of the constructed nomogram model was validated in patients of validation set.Results Braden score,albumin level,alcohol consumption,duration of mechanical ventilation,duration of vasopressor use were identified as independent risk factors for PI in critically ill patients undergoing PPV(all P<0.05).The C-index of the constructed model in the training set and validation set was 0.763(95%CI:0.695-0.831)and 0.732(95%CI:0.596-0.868),respectively.The Calibration curve confirmed good consistency of the model,and theχ2 values of Hosmer-Lemeshow goodness-of-fit tests in the training set and validation set were 10.801 and 8.697(both P>0.05),respectively.Conclusion The constructed risk prediction nomogram model for PI in critically ill patients undergoing PPV demonstrates good predictive performance,which is of value in clinical application.