急性低氧血症性呼吸衰竭患者经鼻高流量氧疗失败风险预测模型的构建

Construction of a risk prediction model for high-flow nasal cannula oxygen therapy failure in patients with acute hypoxemic respiratory failure

目的:探究急性低氧血症性呼吸衰竭患者经鼻高流量氧疗(HFNC)失败的相关因素,并构建其预测模型。方法:回顾性分析2019年1月至2021年7月在中国人民解放军联勤保障部队第908医院接受HFNC治疗的急性低氧血症性呼吸衰竭患者216例,根据HFNC治疗情况,分为成功组(145例)和失败组(转为有创机械通气治疗患者和死亡患者,共71例)。采用单因素和多因素logistic回归分析方法分析HFNC治疗失败的影响因素,并构建风险预测模型。另选取2021年8月至2022年3月于本院接受HFNC治疗的98例急性低氧血症性呼吸衰竭患者作为外部验证,绘制受试者工作特征(ROC)曲线,评估风险预测模型的应用价值,采用Hosmer-Lemeshow检验判断模型的拟合优度。结果:年龄、发热、急性生理学与慢性健康状况评价Ⅱ(APACHEⅡ)评分、冠心病、陈旧性心梗、间质性肺疾病、脑利钠肽(BNP)、心肌肌钙蛋白I(TnI)、初始HFNC流量、患者治疗前后呼吸频率(RR)、ROX指数(SpO2/FiO2与RR的比值)水平在两组之间具有显著差异(P<0.05)。多因素分析结果显示,发热、APACHEⅡ评分、BNP、TnI、治疗1 h后RR、ROX指数均为影响患者HFNC治疗失败的影响因素(P<0.05)。ROC评估结果显示,曲线下面积(AUC)为0.798(95%CI:0.748~0.848),截点值为84.296,灵敏度为81.7%,特异度为63.2%;Hosmer-Lemeshow拟合优度检验显示:该模型预测患者HFNC治疗失败的概率趋近于实际概率(χ^(2)=12.354,P=0.136);以外部数据进行验证,灵敏度为82.1%,特异度为85.7%,总准确率为84.7%。结论:发热、APACHEⅡ评分、BNP、TnI、治疗1 h后RR、ROX指数均为影响患者HFNC治疗失败的影响因素,以此构建风险预测模型的预测效能良好。

Objective:To explore the related factors of failure of high-flow nasal cannula oxygen therapy(HFNC)in patients with acute hypoxemic respiratory failure and construct its prediction model.Methods:A retrospective analysis was performed on 216 patients with acute hypoxemic respiratory failure who received HFNC treatment at the 908 Hospital of the PLA Joint Logistic Support Force from January 2019 to July 2021.Patients were divided into successful group(145 cases)and failure group(71 cases,including patients who were transitioned to receive invasive mechanical ventilation therapy and patients who died)according to the treatment status of HFNC.Univariate and multivariate logistic regression analysis were used to analyze the influencing factors of HFNC treatment failure,and the risk prediction model was constructed.In addition,98 patients with acute hypoxemic respiratory failure who received HFNC treatment in our hospital from August 2021 to March 2022 were selected as external validation.Receiver operating characteristic(ROC)curve was drawn to evaluate the application value of the risk prediction model,and Hosmer-Lemeshow test was used to judge the goodness-fit of the model.Results:There were significant differences in age,fever,acute physiology and chronic health evaluationⅡ(APACHEⅡ)score,coronary heart disease,old myocardial infarction,interstitial lung disease,brain natriuretic peptide(BNP),cardiac troponin I(TnI),initial HFNC flow,and patient's respiratory rate(RR)and ROX index(the ratio of SpO2/FiO2 to RR)before and after treatment between the two groups(P<0.05).Multiva-riate analysis showed that fever,APACHEⅡscore,BNP,TnI,RR and ROX index after 1 hour of treatment were all influencing factors for treatment failure of HFNC(P<0.05).The results of ROC evaluation showed that the area under the curve(AUC)was 0.798(95%CI:0.748-0.848),the cut-off value was 84.296,the sensitivity was 81.7%,and the specificity was 63.2%.Hosmer-Lemeshow goodness of fit test showed that the probability of HFNC treatment failure predicted by this model was close to the actual probability(χ^(2)=12.354,P=0.136).After validating with external data,it showed that the sensitivity was 82.1%,the specificity was 85.7%and the overall accuracy was 84.7%.Conclusion:Fever,APACHEⅡscore,BNP,TnI,RR and ROX index after 1 hour of treatment are all influencing factors for treatment failure of HFNC in patients.Therefore,the predictive efficacy of this risk prediction model is good.