Development And Validation Of A Prognostic Model For Sepsis-induced Acute Respiratory Distress Syndrome

BackgroundAcute respiratory distress syndrome(ARDS)is an acute respiratory failure caused by various factors,manifested as progressive dyspnea and refractory hypoxemia,with rapid onset and high mortality.It is one of the main reasons for mortality in critically ill patients.Sepsis is a life-threatening organ dysfunction due to the dysregulated host response to infection.ARDS occurs in approximately 6%to 7%of patients with sepsis,and the mortality of sepsis-induced ARDS is higher than that of nonsepsis-induced ARDS.Therefore,it is particularly important to investigate the risk factors for the prognosis of sepsis-induced ARDS and identify high-risk patients to improve the prognosis.The commonly used clinical scoring systems for critically ill patients mainly include SOFA and APACHE II,but neither score was constructed based on sepsis-induced ARDS patients.There are certain limitations in application.Currently,there are no scores and prognostic models for predicting the prognosis of patients with sepsis-induced ARDS.ObjectiveThis study aims to establish and validate a short-term prognostic model of sepsis-induced ARDS based on the general information and laboratory examinations of patients,and to provide a reference for early clinical identification of high-risk patients,so as to reduce the mortality of patients.MethodsThe clinical data of sepsis-induced ARDS patients treated in the ICU of the First Affiliated Hospital of Zhengzhou University between October 2017 and October 2021 were retrospectively collected and randomly divided into training cohort and validation cohort according to 7:3.Univariate COX regression,best subsets regression,and Lasso regression were used to initially screen variables,and they were included in multivariate COX regression analysis,and backward stepwise regression was used to screen predictors.The Akaike Information Criterion(AIC)and receiver operating characteristics(ROC)curve evaluated three models.A nomogram model was developed based on the best model,and the C-index,calibration curve and decision curve analysis were analyzed to evaluate the model.Based on the nomogram model,patients were divided into high-risk,intermediate-risk and low-risk groups,and the survival curves of the three groups were drawn,and the Log-Rank test was used to compare the differences between three groups.ROC curve analysis compared the predictive value of the nomogram model,SOFA and APACHE Ⅱ scores in the 28-day prognosis of sepsis-induced ARDS patients.Results1.A total of 462 patients with sepsis-induced ARDS who met the criteria for admission and excretion were enrolled,and the 28-day mortality was 52.60%.2.The variables screened by the best subsets regression was included in the multivariate COX regression analysis,the AIC was the smallest and area under curve(AUC)was largest,and the model fit was better.Multivariate COX regression analysis found that age,NLR>14.56,platelet count<125×109/L,RDW>14.5%,total bilirubin>25μmol/L,albumin<35g/L,and septic shock were associated with 28-day mortality in sepsis-induced ARDS patients.3.The nomogram model based on age,NLR,platelet count,RDW,total bilirubin,albumin and septic shock has good discrimination,accuracy and clinical utility.4.According to the total score,patients were divided into three groups:high-risk(>306.7 points),intermediate-risk(>244.0 points and<306.7 points)and low-risk(≤244.0 points).Kaplan-Meier analysis showed that the survival rate in high-risk group was lower than intermediate-risk and low-risk groups(P<0.01).5.ROC curve analysis showed that in the training cohort,the AUC of the nomogram model was higher than APACHE II and SOFA(Z=4.64,P<0.01;Z=4.61,P<0.01).In the validation cohort,the AUC of the nomogram model was higher than SOFA(Z=2.04,P=0.04)and also higher than APACHE II,but the difference was not statistically significant(Z=1.19,P=0.23).Conclusion1.The nomogram model based on age,NLR,platelet count,RDW,total bilirubin,albumin,combined with septic shock to predict the 28-day mortality in sepsis-induced ARDS patients has good discrimination,accuracy and clinical utility.2.In sepsis-induced ARDS patients,the predictive value of nomogram model for 28-day mortality may be better than the commonly used SOFA and APACHE Ⅱ.