Development And Validation Of A Mortality Prediction Model For Patients With Severe Acute Respiratory Distress Syndrome Receiving Extracorporeal Membrane Oxygenation

BackgroundAcute respiratory distress syndrome is an acute diffuse,inflammatory lung injury that leads to increased pulmonary vascular permeability,increased lung weight,and decreased aerated lung tissue,with high morbidity and mortality.Over the past decades,great progress has been made in the treatment of acute respiratory distress syndrome,including prone ventilation,lung protective ventilation,and lung recruitment.Despite these aggressive treatments,some patients continue to deteriorate and clinicians may consider veno-venous extracorporeal membrane oxygenation.As a salvage therapy,veno-venous extracorporeal membrane oxygenation can provide extracorporeal gas exchange to maintain adequate oxygenation and minimize ventilator-associated lung injury.Advances in extracorporeal membrane oxygenation technology over the last decade have resulted in improved outcomes of patients.However,despite significant advances in this technology,it carries a high risk of complications(e.g.,bleeding,infection,mechanical complications,etc.),results in high mortality,and may expose patients to long-term physical and psychological harm.It is necessary for us to conduct prediction model analysis according to clinical indicators of patients to identify mortality predictors,which can help clinicians to make the best clinical decisions and accurately predict the mortality risk,so we developed this clinical prediction model.ObjectiveTo explore prognostic risk factors in patients with severe acute respiratory distress syndrome receiving extracorporeal membrane oxygenation and develop a risk scoring model to predict the risk of death within 28 days.MethodsIn this multicenter,retrospective study,we enrolled patients with severe acute respiratory distress syndrome who met the inclusion criteria at four study centers:Zhujiang Hospital of Southern Medical University,Shunde Hospital of Southern Medical University,Dongguan People’s Hospital,and Zhongshan People’s Hospital.We collected clinical data before and during extracorporeal membrane oxygenation in these patients,including baseline characteristics,other treatments,and laboratory indicators.The primary outcome was all-cause mortality within 28 days of receiving extracorporeal membrane oxygenation.Logistic regression and Lasso regression were used to screen important prognostic factors and model development,and ROC curves,nomograms,calibration plots,and bootstrap resampling internal validation were used to validate and assess the discrimination and calibration of the prediction model.ResultA total of 190 patients were enrolled in the study,111 were dead within 28 days.After Lasso regression and Logistic regression,8 variables were finally identified as prognostic factors,including:comorbidity,etiology of ARDS,the interval between severe ARDS diagnosis and ECMO initiation,use of glucocorticoids before ECMO,use of vasopressors after initiation of ECMO,Acute Physiology and Chronic Health Evaluation score,platelet level before ECMO,and negative fluid balance on the third day after initiation of ECMO.A prediction model consisting of the above 8 prognostic factors had good discrimination and calibration.The C-statistic was 0.867(95%confidence interval:0.814-0.920)and 0.820 after calibrated by Bootstrap resampling internal validation,suggesting that the model had good discrimination.After Hosmer-Lemeshow goodness-of-fit test,the chi-square value(χ2)was 14.37 and the P value was 0.073,suggesting that the model had good calibration.ConclusionWe successfully constructed a risk score model to predict 28-day mortality in patients with severe acute respiratory distress syndrome receiving extracorporeal membrane oxygenation.It is validated that the model has good discrimination and calibration.Therefore,this nomogram and score model can help clinicians in making optimal medical decisions.