Development and Validation of a Predictive Model for Posttraumatic Hydrocephalus Occurence among Inhospitalized Traumatic Brain Injury Patients – A Single Center Retrospective Study

Introduction: Post-traumatic hydrocephalus(PTH) is an uncommon complication following traumatic brain injury(TBI). This study is to develop and validate a predictive model for PTH occurrence among all inhospitalized TBI patients(both ICU and ordinary ward patients) .

Methods: A retrospective observational study of 2090 admitted TBI patients without previous cerebrospinal fluid shunt operations at Chang Gung Memorial Hospital from January 2015 to December 2016. Split-sample method was used for model derivation(1511 patients) and validation(579 patients). Risk factors were analyzed using multivariable logistic regression(MLR) analysis. C-statistic and Hosmer-Lemeshow statistic was used for model discrimination performance and calibration, respectively.

Results: We identified eight predictive variables：Age (0 point for < 60, 1 point for >/= 60), TBI severity on admission(0 point for GCS 14-15, 1 point for GCS 9-13, 2 points for GCS 3-8), subarachnoid hemorrhage(0 point for no, 1 point for yes), intracerebral hemorrhage (0 point for no, 1 point for yes),intraventricular hemorrhage(0 point for no, 1 point for yes),external ventricular drainage(0 point for no, 3 points for yes), craniotomy(0 point for no, 1 point for yes) and decompressive craniectomy(0 point for no, 1 point for yes). Rates of PTH in the derivation and validation cohorts, respectively, were 0.2% and 0% in extreme low-risk (score = 0~1), 2.1% and 1.5% in low-risk (score = 2-3), 15.5% and 14% in intermediate-risk(score=4-6), 58.8% and 50% in high-risk (score =7~8), both 100% in extreme-high risk(score=9~11) . C-statistic for derivation cohort is 0.88, and for validation cohort 0.92.

Conclusion : This is the first study to the best of our knowledge to propose a predictive model for PTH developmen. The presence of PTH has been proven to be associated with worse long-term outcome, and our PTH predictive model offers physicians an easy and useful tool for early identification and timely treatment of PTH , which is important for faster recovery and lower health expenditure.