基于早期股直肌超声纹理特征的首次有创机械通气患者30天死亡风险预测模型的构建与验证

Establishment and validation of a predictive model for the 30-day mortality risk of patients with the first invasive mechanical ventilation based on the ultrasound texture features of rectus femoris at the early stage

目的基于早期股直肌超声纹理特征,构建首次有创机械通气患者30天死亡风险预测模型及验证。方法采用便利抽样法选择2022年1月至2023年12月入住河南省人民医院呼吸重症患者为研究对象,按8∶2随机分为训练集(n=304)和测试集(n=76)。两组均在插管后第1天、第3天、第5天和第7天使用床旁超声收集纹理特征参数。训练集依据第30天是否死亡为终点事件分为死亡组(n=80)和幸存组(n=224)。采用广义估计模型筛选患者30天死亡的影响因素并建立模型,绘制受试者工作特征(receiver operating characteristic,ROC)曲线下面积(area under receiver operating characteristic curve,AUC)评价模型的区分度,通过校准图和Hosmer-Lemeshow评价模型的校准度。结果年龄、氧合指数(P/F)、患肾脏疾病、使用类固醇激素、股直肌厚度变化率、均值和角二阶距是首次机械通气患者30天死亡的影响因素(P<0.05)。训练集和测试集的AUC分别为0.832(95%CI 0.767~0.898)和0.819(95%CI 0.722~0.916),提示模型的区分度良好。训练集和测试集的Hosmer-Lemeshow分别为5.969(P=0.651)和4.336(P=0.826),校准曲线与对角线重合度较高,说明预测模型的校准度较好。结论本研究基于早期股直肌超声纹理特征建立的预测模型可有效地预测首次机械通气患者30天死亡风险。

Objective To explore the establishment and validation of a predictive model for the 30-day mortality risk of patients with the first invasive mechanical ventilation based on the ultrasound texture features of rectus femoris at the early stage.Methods The patients from the Respiratory Intensive Care Unit of Henan Provincial People′s Hospital by using the convenient sampling method from January 2022 to December 2023.The patients were randomly assigned to the training set(n=304)and the validation set(n=76)in a ratio of 8∶2.The bedside ultrasound was used to collect texture feature parameters on the 1st,3rd,5th and 7th day after intubation.The training set was divided into the death group(n=80)and the survival group(n=224)based on the endpoint event of death or survival on the 30th day.A predictive model for the 30-day mortality risk of patients with the first invasive mechanical ventilation was established by generalized estimating equations.The predictive model was evaluated by the area under receiver operating characteristic(ROC)curve(AUC),calibration curve and Hosmer-Lemeshow.Results Influencing factors of 30-day mortality of patients with the first invasive mechanical ventilation were age,oxygenation index,kidney disease,steroid hormone use,rate of change in the thickness of rectus femoris,mean,and angular second moment(P<0.05).The AUC of the training set and the validation set was 0.832(95%CI 0.767-0.898)and 0.819(95%CI 0.722-0.916),indicating good discrimination of the model.The Hosmer-Lemeshow values of the training set and the validation set were 5.969(P=0.651)and 4.336(P=0.826)respectively.The calibration curve had a high degree of coincidence with the diagonal line,indicating good calibration of the prediction model.Conclusions The model can predict the 30-day mortality risk of patients with the first invasive mechanical ventilation by incorporating ultrasound texture features of rectus femoris at the early stage.