Development And Validation Of A Model For Prediction Of The Risk Of Mechanical Ventilation In Guillain-barré Syndrome

Objectives:The aim of the study is to establish and validate a clinical prediction model for evaluating the risk of mechanical ventilation（MV）in patients with Guillain-Barre syndrome（GBS）.Early prediction of risk of MV in patients with GBS can help clinical workers to furnish patients with optimal management and treatment on admission.Methods:This study retrospectively analyzed 601 patients with GBS who visited our hospital from October 2003 to October 2020.Clinical details from all patients were recorded incorporating sex,age,onset seasons,preceding events,time from onset to admission,bulbar palsy,autonomic dysfunction,tendon reflex on arms and legs at admission,Medical Research Council（MRC）score at admission,C-reactive protein（CRP）,fasting glucose,serum albumin,serum sodium,neutrophils/lymphocyte ratio（NLR）,platelet/lymphocyte ratio（PLR）,and systemic immuneinflammation index（SII）.Patients were separated into MV set and non-MV set based on whether MV was performed during their hospitalization.We randomly split 601 patients with GBS into the training set（400 cases）and the validation set（201 cases）in a 2:1 ratio.In the training cohort,the least absolute shrinkage and selection operator（LASSO）regression model and multivariable logistic regression analysis were applied to establish a predictive model including the predictors selected in LASSO,and nomogram was used to visualize the model.Discriminative ability,calibration,and clinical application of the predictive model were evaluated by utilizing the concordance index（C-index）,calibration curve,and decision curve analysis（DCA）.We externally validated this nomogram based on a validation cohort of 201 GBS patients,and assessed its performance with the C-index,calibration curve,and DCA.Results:1.Our study enrolled 601 eligible patients diagnosed with GBS from 2003 to 2020,including 400 cases（245 males and 155 females）in the training set and 201 cases（117 males and 84 females）in the validation set.The proportion of patients requiring MV in the training and validation sets was 18% and 20%,respectively.2.Univariable analysis in the training set showed MV was significantly related to the following predictive features: age,time from onset to admission,bulbar palsy,autonomic dysfunction,MRC score on admission,serum albumin,fasting glucose,CRP,NLR,PLR,and SII.3.The LASSO regression model selected the optimal predictors,including time from onset to admission,bulbar palsy,autonomic dysfunction,MRC score on admission,fasting glucose,CRP,and NLR.The model predicting the risk of MV was established by using the seven predictors and the model was evaluated and validated.The predictive nomogram demonstrated satisfactory discriminative ability with a C-index of 0.957 in the training group,0.950 in the internal validation group,and 0.917 in the validation group.The calibration curve of the model demonstrated that the prediction probability of the risk of MV in GBS patients was largely close to the actual probability,indicating that the model displayed good calibration.DCA displayed that the model had favorable clinical practical value.Conclusions:1.In our study,univariate and multivariate analysis demonstrated that shorter interval between onset and admission,bulbar palsy,autonomic dysfunction,lower MRC score on admission,elevated fasting glucose,increased CRP,higher neutrophil-to-lymphocyte ratio（NLR）were independent predictive factors for the risk of MV in patients with GBS.2.The clinical predictive nomogram incorporating time from onset to admission,bulbar palsy,autonomic dysfunction,MRC score on admission,fasting glucose,CRP,and NLR showed great discriminative ability and calibration,which can be utilized to assess the possibility of MV in patients with GBS individually.