Establishment And Performance Evaluation Of Early Prediction Model Of Prognosis In Patients With Craniocerebral Injury

Objective:To explore the related factors affecting the prognosis of patients with moderate and severe craniocerebral injury;To establish the early prediction model of prognosis of patients with moderate and severe craniocerebral injury,to predict the prognosis of patients with craniocerebral injury in our hospital(provincial),and to help clinicians and family members of patients to make scientific and reasonable evaluation of prognosis.Methods:To collect the clinical data of 464 cases with moderate and severe craniocerebral injury,which was treated in neurosurgery department of Shanxi provincialpeople’s Hospital from January 2013 to June 2018,including: general condition of patients: age,sex,occupation,cause of injury,type of injury;Clinical severity: GCS score,pupil light reflex,etc;CT scan results: subdural hematoma,epidural hematoma,intracerebral hematoma;laboratory indexes: D-dimer,platelet,etc.The independent risk factors related to prognosis were obtained by univariate and multivariate analysis.The prediction model of prognosis was established by Logistic regression analysis.Internal and external validation model performance,model performance evaluation were appraised by ROC curve under the area,sensitivity,specificity and other indicators.R Software and Empower Stats were used for Statistical Analysis.Results:82 patients died in the early stage of craniocerebral injury,the total fatality rate was17.67%(82/464),the mortality rate was 5.14%(15/292)in 15 patients with moderate craniocerebral injury,and in 67 patients with severe craniocerebral injury,the fatality rate was 38.95%(67/172).There were 51 patients with long-term bad prognosis,the probability of bad prognosis was 24.40%(51/209).multivariate analysis showed that age,pupil light reflex,GCS score,subdural hematoma,epidural hematoma,hemoglobin,platelet,white blood cell are the independent risk factors of prognosis.we establish prediction model(modelA,modelB,modelC)by gradually adding risk factors.The performance of the model increases with the increase of risk factors,and the area under the ROC curve of the model increases from 0.845 to 0.891.Conclusion:1.In this study,it was found that age,low GCS score,disappearance of pupil response to light,epidural hematoma,subdural hematoma,low hemoglobin,low platelet and high white blood cell were the risk factors for predicting the prognosis of patients with moderate and severe craniocerebral injury.2.In this paper,the prediction model of prognosis of patients with craniocerebral injury is established by the above risk factors,and the model of line diagram is generated.The model has good differentiation and accuracy,and has high clinical application value.