Construction And Verification Of A Risk Prediction Model For The Occurrence Of Delayed Cerebral Ischemia After Aneurysmal Subarachnoid Hemorrhage

Objective:Delayed cerebral ischemia(DCI)is a common and serious complication of aneurysmal subarachnoid hemorrhage(a SAH).For a period of time after a patient’s bleeding,if a DCI is complicated,the consequences will be irreversible,so early prevention of its occurrence is particularly important.Based on publicly available data in a public database,we analyzed clinical data of a SAH patients in the neurological intensive care unit of a French hospital,explored risk factors associated with the occurrence of DCI,and established and validated a risk model for the occurrence of DCI after a SAH.Methods:Data on a SAH patients treated in the neurological ICU of a university hospital in France between January 2010 and December 2015 were extracted from the DRYAD database and retrospectively analyzed.A total of 204 patients followed the inclusion and exclusion criteria and were randomly divided into the training group(144 patients)and the verification group(60 patients)in a ratio of 7:3.Based on single-factor and multi-factor analysis,a nomogram model was constructed to verify its feasibility.Receiver operating characteristic curve analysis(Roc curve)was used to test the differentiation of the models.Standard curve and Hosmer Lemeshow(H-L)test were used to determine the calibration degree of the model.Decision curve analysis(DCA)was used to verify the clinical validity of the model.Results:In univariate analysis,differences in external ventricular drainage(EVD),duration of mechanical ventilation,and treatment modalities were statistically significant.Multivariate analysis showed that EVD and rebleeding were significantly correlated with DCI after a SAH.Five clinically relevant features were selected by binary logistic regression to predict the occurrence of DCI in a SAH patients.The area under the operating characteristic curve(AUC)values of the training group and the verification group were 0.768 and 0.754,and the Brier scores were 0.166 and 0.163,respectively.The calibration values of H-L test were x~2=3.824(P=0.923)in the training group and x~2=10.868(P=0.285)in the verification group.The DCA results showed that the training and validation groups showed large positive returns in the broad risk range of 0.77 and 0.63.Conclusion:The predictive model of a SAH combined with DCI has theoretical and practical value,and can provide individualized treatment options for a SAH patients.