Annual Meeting Abstracts: View
(Abst. A.09), 2014
PREDICTIVE MODEL FOR EARLY POST TRAUMATIC SEIZURES IN THE PICU
Authors: Rajsekar Rajaraman, Jason Lerner, Daniel Arndt, David McArthur, Jesse Fischer, Max Zeiger, Meeryo Choe and Christopher Giza
Content: Rationale: Traumatic brain injury (TBI) affects over 500,000 children per year. A large proportion of these patients suffer from early post traumatic seizures (EPTS), defined as a seizure occurring <7 days post-injury. EPTS may indicate a more severe injury or an ongoing process such as an acute hemorrhage, hypoxia, or cerebral edema. Studies report that children are more prone to EPTS than adults, but these studies likely underestimated the EPTS population as there was no continuous EEG (cEEG) monitoring for subclinical seizures. In the study by Arndt and Lerner (2013), possible risk factors for EPTS were identified. In further analyzing the risk factors established in this study cohort, multimodal inference modeling was utilized to identify those with the greatest risk for subclinical seizures and status epilepticus (SE) in pursuit of establishing a predictive model for other institutions. Methods: Consecutive acute TBI patients requiring PICU admission from October 2008-April 2014 at UCLA Medical Center and July 2009 - May 2010 at Children's Hospital of Colorado were consented, placed on cEEG monitoring, and treated with PICU standard of care. All patients had a CT scan during the acute phase. The cohort was evaluated for variables that clinicians typically suspect as possible risk factors for EPTS, including age, gender, Glasgow Coma Scale, clinical seizure in the field, CT scan findings, and mechanism of injury. Statistical analyses were performed using R software (version 2.15.2) (R Core Team ). Results: Of our cohort (n=135), 29% of patients had seizures (39/135), 12.6% having subclinical seizures (17/135), and 11% with subclinical SE (15/135). Abusive head trauma (AHT) was the leading mechanism for subclinical seizures (53.8%). Patients with AHT were 48x more likely to develop subclinical seizures versus other mechanisms (OR, 47.9; 95% CI 11.46-360.89). Additionally, all patients with subclinical seizures had intradural (blood below the dura mater) bleeding on CT scan (17/17). Using multimodal inference modeling, the leading risk factors for subclinical seizures were GCS, young age, and intradural bleed. For subclinical SE, gender was an additional risk factor. Conclusions: Continuous EEG monitoring in pediatric TBI patients significantly decreases morbidity as it is the only method for detecting subclinical seizures, which occurred in 12.6% of this cohort. This study validated the initial Arndt-Lerner study in terms of suspected variables for EPTS. Risk factors, determined by multimodal inference modeling, include GCS, young age, and intradural bleed for subclinical seizures, and additionally, gender for subclinical SE. This data demonstrates the risk for subclinical seizures is greatest for patients <2 years old with intradural blood and/or AHT as the injury of mechanism. Future analyses will include validation of these predictive variables using an independent institutional cohort to create a positive predictive model for EPTS in children. SUPPORTED BY: Thrasher Research Fund, Child Neurology Foundation/Winokur Family Foundation, Today's and Tomorrow's Children Fund, HD061504, NS027544, UCLA BIRC.