Predicting Early Seizures in Children Hospitalized After Mild to Moderate Traumatic Brain Injury

Monday, February 24, 2025

11:05am – 12:05pm Eastern Time

Location: Connections Central Theater 1

Caitlin McNamara, MD (she/her/hers)

Children's Hospital of Pittsburgh of UPMC

Disclosure(s): No relevant financial relationship(s) to disclose.
KV

Kelsey Van Noy, MD

Clinical Instructor
University of Pittsburgh

Disclosure information not submitted.
CP

Christina Patterson, MD

Associate Professor of Pediatrics, Division of Child Neurology
University of Pittsburgh

Disclosure information not submitted.
DS

Dennis Simon, MD

Childrens Hospital of Pittsburgh of UPMC

Disclosure(s): No relevant financial relationship(s) to disclose.
RB

Rachel Berger, MD, MPH

Director, Child Abuse Research, Safar Center for Resuscitation Research, University of Pittsburgh
University of Pittsburgh

Disclosure information not submitted.
Patrick M. Kochanek, MD, MCCM

Distinguished Professor of Critical Care Medicine
UPMC Presbyterian

Disclosure information not submitted.
EF

Ericka Fink, MD, FCCM

Professor of Critical Care Medicine, Pediatrics, and Clinical and Translational Science
Childrens Hospital of Pittsburgh of UPMC

Disclosure information not submitted.
CH

Christopher Horvat, MD, MHA

Childrens Hospital of Pittsburgh of UPMC

Disclosure information not submitted.

Disclosure(s):

Caitlin McNamara, MD: No relevant financial relationship(s) to disclose.

Dennis Simon, MD: No relevant financial relationship(s) to disclose.

Introduction: Seizures are common among infants and toddlers with mild-moderate traumatic brain injury (TBI) and are associated with worse outcomes. We aimed to create a risk prediction model for early seizures in this population.

Methods: Patients < 3 years old (y) admitted to our pediatric intensive care unit after mild to moderate TBI (initial GCS 9-15) between 2011-2021 were included. The outcome was early electrographic or clinical seizures within the first 3 days of admission. A multivariable model incorporated predictors with P< 0.05 in univariable analysis. Internal 5-fold cross validation was used to assess performance measured by area under the receiver operating characteristic curve (AUROC) and area under the precision recall curve (AUPRC). Thresholds for continuous variables were chosen using locally estimated scatterplot smoothing. Coefficients were rounded to the nearest integer to create a point system to facilitate bedside model risk calculation.

Results: Six-hundred thirty-three patients met inclusion criteria with 40 (6%) having early seizures. There were low rates of EEG (149/633 (24%)) and patients who received seizure prophylaxis (251/633 (40%)). A 9 variable prediction model was developed that included (points): caregiver reported seizure at presentation (+2), initial prothrombin time ( 15 sec, +2), age  12 months (+1.5), subdural hemorrhage (+1), intraparenchymal hemorrhage (+1), initial glucose (  200 mg/dL, +0.5), epidural hemorrhage (-0.5), no intracranial hemorrhage (-1), and initial hemoglobin > 7.5 g/dL (-2). AUROC was 0.88 (95% confidence interval 0.82-0.94) and AUPRC was 0.41. Goodness of fit test showed acceptable calibration (P=0.9). Point thresholds of 1 for administering anti-seizure prophylaxis and 2.5 for obtaining an EEG would miss 10% and 28% of patients who developed early seizures, respectively.

Conclusions: This predictive model using 9 variables in pediatric patients < 3 y with mild to moderate TBI could help identify who is at high risk for early seizures and who should receive seizure prophylaxis or EEGs (to detect subclinical seizures despite prophylaxis). Both reported thresholds are low, consistent with the high risk for post-traumatic seizures in this patient population. This model merits additional validation and refinement.