| Objective: Traumatic brain injury remains the leading cause of both morbidity and mortality in children.Fortunately,most head injuries in children are classified as mild traumatic brain injury(m TBI).Most children recover from these injuries uneventfully when treated appropriately.However,in some patients,m TBI can be associated with significant intracranial injury or progressive hemorrhage injury(PHI),including developmental and psychosocial delays in addition to long-term disability or death.For the diagnosis of m TBI children with intracranial injury or progressive hemorrhage injury,clinicians often chose frequent cranial computed tomography(CT)scans.However,the utility of repeated CT scans,especially in children with mild TBI,has been debated.Children are more vulnerable to the carcinogenic effects of radiation than adults.Minimizing the number of CT scans is of great importance in children,because of the increased risk of malignancies induced by exposure to ionizing radiation.Thus,retrospectively collected data available at admission were used to develop two prognostic models and determine the likelihood of intracranial injury and PHI after pediatric m TBI.Finally,the objectives of this study were to produce a certain guiding significance for the classification,monitoring and rescue of the disease,and provide the theoretical basis for the standardized application of CT scan in children.Methods: Between 2014 and 2016,we retrospectively enrolled infants <2 years old with mild TBI.Data recorded included age,sex,mechanism of head injury,scalp hematoma,vomiting and intracranial injuries on CT.And we also retrospectively enrolled children <15years of age with mild TBI and recorded age,sex,GCS scores on admission,causes of head injury,timing of initial CT,any loss of consciousness,vomiting and seizure data,and type of TBI.The collected data were used to develop intracranial injury and PHI predictive models with univariate analysis and multivariate logistic regression analysis.Independent test data sets were used to verify the models and to observe the performances of the prediction models.Finally,the prediction tools were developed with the aid of programming technology.Results: Of 214 enrolled infants <2 years old,30(5.8%)sustained intracranial injuries.Univariate analysis revealed that age in month,the injury mechanism,scalp hematoma,and vomiting times were associated with intracranial injuries on CT.After FDR conversion,the injury mechanism,scalp hematoma,and vomiting times were selected into the multiple factors analysis.Severe injury mechanism,vomiting times and scalp hematomas were independently associated with traumatic intracranial injuries.The C-statistic for the intracranial injury model is 0.949,the best cutoff P value is 0.132.When the P value of the infant is greater than 0.132,it is believed that the possibility of intracranial injury is higher.The validation of test dataset showed that MCC value is 0.670 and F-measure value is 0.780.Of the 175 enrolled children<15years,15(8.6%)experienced PHI.Univariate analysis revealed that GCS score on admission,vomiting,seizure,and TBI type were associated with PHI.After FDR conversion,GCS score on admission,vomiting,seizure,and TBI type were selected into the multiple factors analysis.Multivariate logistic regression analysis showed that GCS score on admission,epidural hemorrhage and vomiting were independently associated with PHI.The C-statistic for the PHI model is 0.789,the best cutoff P value is 0.100.When the P value of the children is greater than 0.100,it is believed that the possibility of PHI is higher.The validation of test set showed that MCC value is 0.741 and F-measure value is 0.800.Conclusion: The established predictive models may conveniently and accurately predict the occurrence of intracranial injury and progressive hemorrhagic injury after pediatric m TBI in the early stage,and the developed prediction tools may help in the decision making in clinics. |
PDF Full Text Request