Sage Journals: Discover world-class research

Abstract

Early seizures are common among infants and toddlers hospitalized with mild-moderate traumatic brain injury (TBI) and are associated with unfavorable outcomes. We aimed to develop a risk prediction model for early seizures in this population to support clinical care. Patients <3 years old admitted to our pediatric intensive care unit after mild to moderate TBI (initial Glasgow Coma Scale score 9–15) between 2011 and 2024 were included. The outcome was early electrographic or clinical seizures within the first 3 days of admission. Variable selection was performed using Lasso regression. Variables selected by Lasso were subsequently included in a multivariable logistic regression model, trained using repeated fivefold cross-validation. Model discrimination was assessed using area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC). Calibration was evaluated with Hosmer–Lemeshow tests and reliability plots. 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. All values are reported with 95% confidence intervals in parentheses. Of 796 patients who met inclusion criteria, 52 (6%) had early seizures, electroencephalography (EEG) was obtained in 180 (23%), and antiseizure medication prophylaxis was administered to 299 (38%). A 9 variable, point-based prediction model was developed that included (points): age <12 months (+2), presentation concerning for seizure (+2), bruising on physical exam (+0.5), subdural hemorrhage (+1.5), intraparenchymal hemorrhage (+1.5), initial hemoglobin ≤ 7.5 g/dL (+1.5), initial prothrombin time ≥ 15 sec (+1.5), sodium ≥ 145 mmol/L (+2), and glucose ≥ 200 mg/dL (+0.5). AUROC was 0.89 (0.84, 0.95), and AUPRC was 0.46 (0.32, 0.59). The goodness-of-fit test showed acceptable calibration (p = 0.22), as did visualization of a reliability plot. Point thresholds of 1 and 3 had sensitivities of 0.96 (0.87, 0.99) and 0.65 (0.51, 0.78) with specificities of 0.44 (0.40, 0.47) and 0.91 (0.89, 0.93), respectively. This predictive model using nine variables in pediatric patients less than 3 years old with mild to moderate TBI could help identify patients at high risk for early seizures and who should receive seizure prophylaxis or EEG monitoring. This model merits additional validation and refinement.

Keywords

child abuse pediatric predictive modeling seizure traumatic brain injury

Introduction

Seizures are common in young children admitted with traumatic brain injury (TBI) of any severity, occurring in 2–48% of these patients.^1–6 Early post-traumatic seizures (EPTS), defined as occurring within the first 3 days after TBI is incurred, have been shown to be preventable with antiseizure drug prophylaxis, but there are no guidelines on prophylaxis or other management for children with mild to moderate pediatric TBI.^7–9 Risk factors for EPTS include younger age, higher TBI severity (as defined by lower Glasgow Coma Scale [GCS]), and abusive head trauma (AHT) etiology.^2,3 AHT is defined as TBI due to physical abuse, which includes shaking or blunt impact, in patients younger than 5 years old.^10–13 Since EPTS are shown to be associated with worse outcomes as well, early recognition and treatment could improve outcomes in this vulnerable population.^14,15

Antiseizure drug prophylaxis is recommended in published guidelines on severe pediatric TBI (presenting GCS ≤ 8), but there are no guidelines for seizure prophylaxis for mild to moderate pediatric TBI (presenting GCS 9–15).⁸ Certain postinjury findings, such as subdural hemorrhages or ischemia, are known risk factors for seizures, but there is wide variation in antiseizure drug prophylaxis practice.^16–19 While antiseizure drug prophylaxis reduces EPTS, the literature is mixed on the drug of choice, contributing to the prescribing variability.^20–23 There are also potential risks to antiseizure drug prophylaxis that include side effects such as psychobehavioral cognitive effects, motor instability/ataxia, and concerns for the medications contributing to long-term neurodevelopmental delay, as well as costs to the patients.^24–28 Among children with AHT, individualized antiseizure drug prophylaxis may help to prevent seizures while avoiding unnecessary medication initiation in children who are low risk.²⁹ As EPTS are associated with worse outcomes, can be clinically challenging to detect, and require resource-intensive electroencephalography (EEG) to confirm, a well-constructed predictive model to characterize risk for EPTS is a promising approach to accurately identify more at-risk patients while continuing to promote appropriate stewardship of both EEG and antiseizure drug prophylaxis.^29–31

In pediatric patients with mild or moderate TBI, as defined by an admission GCS score of 9–15, the decision to obtain an EEG can be challenging. While some children have clearly abnormal neurological exams that warrant EEG monitoring, many appear neurologically intact despite serious injury. In addition, EEG monitoring is a resource-intensive tool that requires technicians, EEG machines, and neurologists, which are not always readily available.^32–34 This limitation makes clinical stratification important to ensure patients obtain the appropriate services. In AHT, nonconvulsive seizures are common, affecting 57–65% of these patients.^35–37 A 2015 article by O’Neill and colleagues found that 26% of their pediatric TBI cohort presented with a mild or moderate GCS score, and 74% of the patients who seized were diagnosed with AHT.³⁷ Given limited EEG resources and the high seizure risk in some seemingly well-appearing children, better tools are needed to identify those who would benefit most from EEG monitoring.

Our objective was to create a prediction model to assess the risk of EPTS in patients under 3 years old with mild to moderate TBI, a population that includes a large proportion of AHT. The intended purpose of such a model is to detect occult seizure risk, which could assist clinicians in deciding who should receive antiseizure drug prophylaxis and/or EEGs.

Methods

The University of Pittsburgh Institutional Review Board (IRB) approved this retrospective chart review (STUDY21060042). Procedures were followed in accordance with the Helsinki Declaration of 1975. Patients were included if they were under 3 years old, admitted to the UPMC Children’s Hospital of Pittsburgh pediatric intensive care unit (PICU) between January 1, 2011, and July 31, 2024, with a diagnosis of TBI and a presenting GCS score above 8. The GCS score was the first obtained upon presentation to the emergency department. Mild GCS was defined as a score between 13 and 15 and moderate between 9 and 12. Patients were excluded if they had penetrating head trauma.

Variables abstracted from the electronic health record (EHR) included age, TBI etiology (AHT vs. accidental TBI [aTBI]), history of presenting illness, physical exam findings, type of intracranial hemorrhage, EEG results, presence of antiseizure drug prophylaxis administration, initial laboratory values (hemoglobin [g/dL], platelets [×10e⁹/L], prothrombin time [sec], partial thromboplastin time [sec], international normalized ratio [INR], fibrinogen [mg/dL], sodium [mMol/L], glucose [mg/dL], blood urea nitrogen [mg/dL], creatinine [mg/dL]), and admission GCS score. An independent radiology review was not carried out, but at our institution, a pediatric neuroradiology attending reads all the neurological images, including those obtained at another hospital.

The primary outcome was the presence of EPTS, defined as documented clinical or electrographic seizures within the first 3 days of admission. Children were classified as having AHT if the Child Protection Team determined the TBI was “highly concerning” or “diagnostic” for AHT. Other blunt TBI etiologies were classified as aTBI. In the history of presenting illness documented in the EMR, a patient was characterized as having a presentation concerning for seizure if a caregiver described an episode concerning for a seizure prior to medical care. Antiseizure drug prophylaxis was defined as any antiseizure drug given before seizure occurrence (clinical or EEG). During this time period, there were no institutional guidelines for mild to moderate TBI regarding which patients to give antiseizure drug prophylaxis or to obtain an EEG. Thus, both giving antiseizure drug prophylaxis and obtaining an EEG were up to the clinician discretion.

Missing laboratory data (<3%) were imputed. Nonparametric tests, including Wilcoxon rank-sum test, Pearson’s chi-square test, and Fisher’s exact test, as appropriate, were used to compare characteristics between patients with and without EPTS. Lasso regression to predict EPTS was conducted to choose variables that would be available at admission. In the Lasso regression, the following variables were inputted: age in months, presentation concerning for seizure, bruising on physical exam, epidural hemorrhage, subdural hemorrhage, subarachnoid hemorrhage, intraparenchymal hemorrhage, no hemorrhage, sodium, glucose, blood urea nitrogen, creatinine, hemoglobin, platelet count, prothrombin time, partial thromboplastin time, and INR.

Thresholds for continuous variables were initially explored using locally estimated scatterplot smoothing (LOESS) to visualize nonlinear associations with EPTS. Candidate thresholds suggested by these visualizations were then evaluated using iterative multivariable logistic regression using repeated k-fold internal cross-validation (fivefolds). The variable correlation coefficients were then rounded to the nearest half point to assign a point value for each variable. Variable points were then added together for each patient to assign a summed point value. The area under receiver operator curve (AUROC) and area under precision-recall curve (AUPRC) were calculated. Goodness of fit was assessed with the Hosmer–Lemeshow test and visualization of reliability plots. All statistical analyses were performed by using RStudio Version 2024.12.1 + 563 (RStudio, Boston, MA) and R Version 4.4.3 (R Foundation for Statistical Computing, Vienna, Austria) with the following packages: readxl, tidyverse, gtsummary, lubridate, glmnet, and mice.

Results

Of 995 patients admitted to the PICU less than 3 years old with TBI, 39 were excluded for penetrating TBI and 160 were excluded for severe TBI (Fig. 1). There were 796 patients less than 3 years old with mild or moderate TBI. The median age of the cohort was 4 months (interquartile range [IQR] 2, 14), with 38% female and 33% with AHT. Most patients were healthy prior to the TBI (83%). The most common past medical history was prematurity (9%). One patient had a history of a seizure, and two patients had a history of TBI. During hospitalization, 65/796 (8%) had seizures, and 52/65 (80%) were classified as EPTS. In the entire cohort, 299/796 (38%) had antiseizure drug prophylaxis given, and 167/796 patients (21%) had EEGs obtained. Of the patients who had seizures, 15/65 (23%) had clinical seizures, and 51/65 (78%) were diagnosed with AHT. There was no difference in the age between the patients with and without seizures. There were 99 patients with a presentation concerning for seizure, 34 (34%) of whom had a seizure. All patients had neuroimaging, with 492/796 (62%) having head CTs only, 49/796 (6%) having brain MRIs only, and 255/796 (32%) having both head CT and brain MRI. The most common type of intracranial hemorrhage was subdural, followed by subarachnoid, epidural, intraparenchymal, and intraventricular hemorrhages (Table 1). In the 65 patients who seized, most had subdural hemorrhage, and only one had no hemorrhage. Antiseizure drug prophylaxis was used in 304/796 (37%) patients, and all patients who seized were given antiseizure drug prophylaxis (Table 1). The median laboratory values that were significantly different between the patients with EPTS versus those with no seizures were hemoglobin, prothrombin time, INR, and glucose.

FIG. 1.

Patient flowchart shows the flowchart of patients in terms of seizures and seizure timing.

Table 1.

Patient Characteristics by Seizure Timing

	Overall	Early	None	p value^b
Characteristic	N = 783^a	N = 52^a	N = 731^a	p value^b
Age (months)	4 (2, 14)	3 (2, 6)	5 (2, 14)	0.04
TBI etiology: abusive head trauma	259 (33%)	43 (83%)	216 (30%)	<0.01
Presentation concerning for seizure	93 (12%)	28 (54%)	65 (9%)	<0.01
Antiseizure drug prophylaxis	286 (37%)	52 (100%)	234 (32%)	<0.01
EEG obtained	167 (21%)	51 (98%)	116 (16%)	<0.01
Imaging findings
Epidural hemorrhage	136 (17%)	3 (6%)	133 (18%)	0.02
Subdural hemorrhage	429 (55%)	48 (92%)	381 (52%)	<0.01
Subarachnoid hemorrhage	223 (28%)	18 (35%)	205 (28%)	0.31
Intraparenchymal hemorrhage	34 (4%)	10 (19%)	24 (3%)	<0.01
Intraventricular hemorrhage	11 (1%)	2 (4%)	9 (1%)	0.16
No hemorrhage	115 (15%)	1 (2%)	114 (16%)	<0.01
Laboratory values
Hemoglobin [g/dL]	11.3 (10.3, 12.1)	9.8 (8.8, 10.6)	11.4 (10.5, 12.2)	<0.01
Platelet [×10e⁹/L]	395 (322, 483)	424 (328, 526)	394 (322, 479)	0.21
Prothrombin time (PT) [sec]	13.6 (13.0, 14.3)	14.5 (14.0, 15.4)	13.6 (13.0, 14.2)	<0.01
Partial thromboplastin time (PTT) [sec]	31 (28, 34)	31 (29, 34)	31 (28, 34)	0.97
International normalized ratio (INR)	1.1 (1.0, 1.1)	1.1 (1.1, 1.2)	1.0 (1.0, 1.1)	<0.01
Fibrinogen [mg/dL]	253 (211, 302)	286 (218, 334)	252 (210, 299)	0.17
Sodium [mMol/L]	138 (137, 140)	139 (137, 141)	138 (137, 140)	0.04
Glucose [mg/dL]	96 (86, 111)	116 (92, 148)	95 (86, 110)	<0.01
Blood urea nitrogen (BUN) [mg/dL]	10 (8, 13)	10 (7, 13)	10 (8, 13)	0.95
Creatinine (Cr) [mg/dL]	0.20 (0.17, 0.30)	0.20 (0.18, 0.28)	0.20 (0.17, 0.30)	0.46

Median (Q1, Q3); n (%).

Kruskal–Wallis rank sum test; Fisher’s exact test.

EEG, electroencephalography; TBI, traumatic brain injury.

The following variables had significant coefficients in the Lasso regression: age in months, presentation concerning for seizure, bruising, subdural hemorrhage, intraparenchymal hemorrhage, initial hemoglobin, initial prothrombin time, initial sodium, and initial glucose (Table 2). For the continuous variables, LOESS was conducted for each and found thresholds for predictability of EPTS: age <12 months, hemoglobin ≤ 7.5 g/dL, prothrombin time ≥ 15 sec, sodium ≥ 145 mMol/L, and glucose ≥ 200 mg/dL (Fig. 2).

FIG. 2.

Locally estimated scatterplot smoothing (LOESS) for continuous variable thresholds. For glucose (A), partial thromboplastin time (B), hemoglobin (C), sodium (D), prothrombin time (E), and age (F), LOESS was utilized to find probability thresholds for predicting early seizures. These thresholds were then used in the predictive model.

Table 2.

Model Creation

	LASSO coefficient	Correlation coefficient	Scaled points
Age <12 months	0.045	1.85	2.0
Presentation concerning for seizure	0.197	2.13	2.0
Bruising on physical exam	0.001	0.34	0.5
Subdural hemorrhage	0.034	1.41	1.5
Intraparenchymal hemorrhage	0.153	1.70	1.5
Hemoglobin ≤ 7.5 mg/dL	0.116	1.68	1.5
Prothrombin time ≥ 14 sec	0.111	1.58	1.5
Sodium ≥ 145 mmol/L	0.253	1.87	2.0
Glucose ≥ 200 mg/dL	0.037	0.69	0.5

After scaling the variable correlation coefficients, the points assigned for each variable were as follows: age <12 months (+2), presentation concerning for seizure (+2), bruising (+0.5), subdural hemorrhage (+1.5), intraparenchymal hemorrhage (+1.5), hemoglobin ≤ 7.5 g/dL (+1.5), prothrombin time ≥ 15 sec (+1.5), sodium ≥ 145 mmol/L (+2), and glucose ≥ 200 mg/dL (+0.5). The AUROC was 0.89 (0.84, 0.95), and the AUPRC was 0.46 (0.32, 0.59) (Fig. 3). The Hosmer–Lemeshow goodness-of-fit test was not significant (Fig. 3). In the Hosmer–Lemeshow plot, predicted probabilities were skewed, with a median value of 0.036 (IQR: 0.008–0.036), and most predictions clustered below 0.05.

FIG. 3.

Model Performance. (A) The area under the receiver operating curve (AUROC) for the early seizure predictive model. (B) The area under the precision-recall curve (AUPRC) for the early seizure predictive model. (C) The goodness-of-fit test conducted with the Hosmer–Lemeshow test. The observed probabilities are on the y-axis, and the expected probabilities are on the x-axis. (A) Receiver Operating Curve. (B) Precision Recall Curve. (C) Hosmer–Lemeshow test.

The average points for patients with EPTS were 4.0 (2.0, 4.0) versus 2.0 (0.0, 2.0) in the patients without EPTS (p < 0.01) (Table 3). The reliability plot is available in Supplementary Figure S1. Most patients with EPTS (34/52 [65%]) had 3 points or higher, while most patients without EPTS had less than 3 points (679/744 [91%]). Of the patients with points greater than or equal to 2, 147/468 (31%) obtained an EEG and 232/468 (50%) were given antiseizure drug prophylaxis. Of the patients with points greater than or equal to 3, 83/99 (84%) obtained an EEG, and 91/99 (92%) were given antiseizure drug prophylaxis.

Table 3.

Point System by Seizure

	Overall	No seizure	Seizure
	N = 796	N = 744	N = 52
0–0.9	328 (41%)	326 (44%)	2 (4%)
1–1.9	224 (28%)	219 (29%)	5 (10%)
2–2.9	145 (18%)	134 (18%)	11 (21%)
3–3.9	10 (1%)	4 (1%)	6 (12%)
4–4.9	81 (10%)	55 (7%)	26 (50%)
>5.0	8 (1%)	6 (1%)	2 (4%)

Median (Q1, Q3); n (%).

At a threshold of 1 point in the final model, the sensitivity was 0.96 (0.87, 0.99) and the specificity was 0.44 (0.40, 0.47) with a positive predictive value (PPV) of 0.11 (0.99, 0.12) and a negative predictive value (NPV) of 0.99 (0.98, 0.99). The 1-point threshold would miss 2 patients out of 52 (4%) with EPTS, and there would be 418 false positives. At a threshold of 3 points, the sensitivity was 0.65 (0.51, 0.78) and the specificity was 0.91 (0.89, 0.93) with a PPV of 0.34 (0.28, 0.42) and an NPV of 0.97 (0.96, 0.98). The 3-point threshold would miss 18/52 (34.6%) with EPTS, and there would be 65 false positives.

Discussion

Using more than a decade of data from a quaternary children’s hospital, we developed a high-performing prediction model to identify children with TBI who are at risk for EPTS. Prioritizing parsimony, we ultimately relied on six data elements routinely collected during routine TBI care, enabling straightforward adoption at the bedside. When applied as a point-based system, the model demonstrated strong overall performance, with threshold-based scoring yielding high specificity and NPV.

Our model focuses on patients under 3 years old, a population in which AHT is a leading consideration in the differential diagnosis of TBI and the risk for seizures is correspondingly high.^3,4,37–39 We intentionally excluded AHT diagnosis itself as a predictor, as it is rarely established at the time EEG monitoring or seizure prophylaxis decisions are made.^40–43 Instead, we selected variables that are both easily obtainable and available at admission, several of which, such as low hemoglobin, bruising, and subdural hematoma, are strongly associated with AHT.^11,41,44 Two primary models for predicting AHT exist: PredAHT-2, which incorporates variables from the entire hospitalization, and PediBIRN, which relies on physical exam findings and neuroimaging.^45,46 In contrast, our model integrates age, presenting symptoms, laboratory results, and neuroimaging findings, allowing application across a broader range of injury mechanisms and including cases of milder injury severity.

We believe this model can be used as a risk stratification tool to help with clinical decision-making. Lower scores identify patients at low risk for EPTS, while progressively higher scores correspond to increasing seizure risk. A higher score justifies intervention with antiseizure drug prophylaxis or monitoring with EEG, while a patient with a low score may not require anything beyond observation. For example, a 12-month-old infant admitted after a fall with a GCS score of 15, bruising on physical examination, normal laboratory values, and a subdural hemorrhage on imaging would score 2 points, placing them in an intermediate risk category. In this scenario, antiseizure drug prophylaxis without EEG monitoring would be reasonable to prevent EPTS. A 4-month-old presenting with a presentation concerning for seizure, subdural hemorrhage, and a hemoglobin of 7 mg/dL would score 7 points and be considered very high risk for EPTS, warranting both EEG monitoring and antiseizure drug prophylaxis. Further studies are warranted to define thresholds for EEG monitoring and antiseizure drug prophylaxis.

There are no guidelines for antiseizure drug prophylaxis for mild to moderate TBI, and although certain risk factors are recognized, clinical practice varies widely, with reported utilization rates ranging from 9% to 52%.^47–52 The most efficacious antiseizure drug prophylaxis is also undefined, making universal prophylaxis challenging given the potential for adverse effects.^{7,9,20,51,53,54} Older agents, such as fosphenytoin, are associated with ataxia, liver dysfunction, or drowsiness, while the current first-line agent, levetiracetam, commonly causes irritability.^55–57 Moreover, prolonged or unnecessary prophylaxis may impair TBI recovery.^26,53 Further research is needed to determine optimal prophylaxis strategies, as factors such as drug selection, dosing, and injury endotype likely influence therapeutic efficacy.^20,23,58 In our cohort, antiseizure drug prophylaxis was often administered without EEG monitoring, potentially reflecting scenarios where the neurological exam was reassuring but neuroimaging findings were concerning.

In TBI, including AHT, seizures are associated with secondary injury and worse outcomes.^16,29,31 The true prevalence of EPTS is uncertain, as many events are nonconvulsive and EEG is not routinely performed in all patients.^4,35,36 Based on our model’s thresholds and performance metrics, this tool may aid clinicians in deciding when to obtain an EEG. Key questions arising from our findings include whether the model could guide not only which patients might benefit from EEG but also, given the risk of subclinical status epilepticus, which patients might warrant continuous EEG monitoring. Future investigation is needed, as continuous EEG is resource-intensive and often limited in availability. Current epilepsy guidelines recommend 24/7 real-time supervision by trained personnel with direct access to EEG technologists, nurses, and physicians.^59,60 However, a national shortage of nursing and EEG technologists has been documented, and limited or delayed EEG access has been shown to postpone care.^60,61 Given that longer seizure duration reduces responsiveness to anticonvulsant therapy and worsens outcomes, prioritizing EEG resources for higher risk patients in this model could be clinically beneficial.^62–66 This would target a population who would likely be given antiseizure drug prophylaxis and thus would be more likely to have unrecognized subclinical seizures.

Our study has several limitations. First, the data are from a single center and may not reflect broader clinical practice. Second, the retrospective design raises the possibility that clinical seizures were missed due to incomplete documentation in the EHR or misclassification of antiseizure drug administration as prophylaxis versus treatment. A large proportion of our population received antiseizure drug prophylaxis, which could have altered the rate of seizures. In addition, all patients who had early seizures received antiseizure drug prophylaxis, which could be misclassified as prophylaxis, a signal that prophylaxis is ineffective, and/or that clinicians have a good sense of seizure risk in these patients that was not otherwise documented. Because EEGs were not routinely obtained, nonconvulsive seizures may have gone undetected. We also did not go into detail regarding the number or type of seizure for the patients who did seize or the antiseizure drugs the patients received. Our model also did not leverage the full range of predicted probabilities, which may limit its utility in stratifying patients into distinct high- versus low-risk categories. Finally, although a rationale for model thresholds was provided, these thresholds need to be studied further to determine if they are optimal. Despite these limitations, our model demonstrated excellent predictive performance. Future work should refine using multicenter data and evaluate its utility prospectively.

Conclusion

We developed a predictive model for EPTS in patients under 3 years old admitted to the PICU with mild to moderate TBI. The model demonstrates how to risk-stratify and allows for clinicians to make educated decisions about antiseizure drug prophylaxis and EEG monitoring, with the aim of preventing seizures and enabling early detection and treatment to improve outcomes. Prospective validation will be essential before clinical implementation.

Transparency, Rigor, and Reproducibility Summary

This retrospective cohort study was conducted at a single center and included consecutive patients under 3 years old admitted with mild to moderate TBI from 2011 to 2024. All inclusion and exclusion criteria, as well as outcome definitions, were prespecified and defined in the Methods. STROBE guidelines for reporting observational studies and TRIPOD guidelines for predictive model development were followed. Data were extracted from the EHR using standardized definitions. Due to patient privacy considerations and institutional IRB restrictions, the full dataset cannot be made publicly available. Deidentified summary data and analytic code are available upon request. We used Lasso regression for variable selection followed by multivariable logistic regression to build the predictive model. Thresholds for continuous predictors were determined using LOESS smoothing. Model performance was evaluated using internal 5-fold cross-validation followed by AUROC and AUPRC. Calibration was assessed using the Hosmer-Lemeshow test and reliability plots. This study was approved by the University of Pittsburgh Institutional Review Board (STUDY21060042) and conducted in accordance with the Declaration of Helsinki. Informed consent was waived for this analysis.

Authors’ Contributions

C.R.M. had full access to all the data in the study and takes responsibility for the integrity of the data and accuracy of the data analysis. Concept and design: C.R.M. and C.M.H. Acquisition, analysis, or interpretation of data: All authors. Drafting of the article: C.R.M., P.M.K., and C.M.H. Critical revision of the article for important intellectual content: all authors. Statistical analysis: C.R.M.

Footnotes

Supplemental Material

References

1. Elsamadicy

, Koo

, David

, et al. Post-traumatic seizures following pediatric traumatic brain injury. Clin Neurol Neurosurg 2021;203:106556; doi: 10.1016/j.clineuro.2021.106556

2. Chung

, O’Brien

. Prevalence of early posttraumatic seizures in children with moderate to severe traumatic brain injury despite levetiracetam prophylaxis. Pediatr Crit Care Med 2016;17(2):150–156; doi: 10.1097/PCC.0000000000000588

3. Liesemer

, Bratton

, Zebrack

, et al. Early post-traumatic seizures in moderate to severe pediatric traumatic brain injury: Rates, risk factors, and clinical features. J Neurotrauma 2011;28(5):755–762; doi: 10.1089/neu.2010.1518

4. Arndt

, Lerner

, Matsumoto

, et al. Subclinical early posttraumatic seizures detected by continuous EEG monitoring in a consecutive pediatric cohort. Epilepsia 2013;54(10):1780–1788; doi: 10.1111/epi.12369

5. Rumalla

, Smith

, Letchuman

, et al. Nationwide incidence and risk factors for posttraumatic seizures in children with traumatic brain injury. J Neurosurg Pediatr 2018;22(6):684–693; doi: 10.3171/2018.6.PEDS1813

6. Bennett

, DeWitt

, Harlaar

, et al. Seizures in children with severe traumatic brain injury*. Pediatr Crit Care Med 2017;18(1):54–63; doi: 10.1097/PCC.0000000000000948

7. Samara

, Ifraitekh

, Al Jayyousi

, et al. Use of antiepileptic drugs as prophylaxis against posttraumatic seizures in the pediatric population: A systematic review and meta-analysis. Neurosurg Rev 2023;46(1):49; doi: 10.1007/s10143-023-01963-z

8. Kochanek

, Tasker

, Carney

, et al. Guidelines for the management of pediatric severe traumatic brain injury, third edition: Update of the brain trauma foundation guidelines . Pediatr Crit Care Med 2019;20(3):280–289; doi: 10.1097/PCC.0000000000001735

9. Kolf

, McPherson

, Kniska

, et al. Early post-traumatic seizure occurrence in pediatric patients receiving levetiracetam prophylaxis with severe traumatic brain injury. J Pediatr Pharmacol Ther 2020;25(3):241–245; doi: 10.5863/1551-6776-25.3.241

10.

10. Jenny

, Hymel

, Ritzen

, et al. Analysis of missed cases of abusive head trauma. Journal of the American Medical Association 1999;281(7):621–626; doi: 10.1001/jama.281.7.621

11.

11. Berger

, Fromkin

, Herman

, et al. Validation of the Pittsburgh infant brain injury score for abusive head trauma. Pediatrics 2016;138(1):e20153756; doi: 10.1542/peds.2015-3756

12.

12. Sheets

, Leach

, Koszewski

, et al. Sentinel injuries in infants evaluated for child physical abuse. Pediatrics 2013;131(4):701–707; doi: 10.1542/peds.2012-2780

13.

13. Parks

, Annest

, Hill

, et al. Pediatric abusive head trauma: Recommended definitions for public health surveillance and research. Centers for Disease Control and Prevention 2012.

14.

14. McNamara

, Kalinowski

, Horvat

, et al. New functional impairment after hospital discharge by traumatic brain injury mechanism in younger than 3 years old admitted to the PICU in a single center retrospective study. Pediatr Crit Care Med 2024;25(3):250–258; doi: 10.1097/PCC.0000000000003417

15.

15. Keenan

, Clark

, Holubkov

, et al. Longitudinal developmental outcomes of infants and toddlers with traumatic brain injury. JAMA Netw Open 2023;6(1):e2251195; doi: 10.1001/jamanetworkopen.2022.51195

16.

16. Dingman

, Stence

, O’Neill

, et al. Seizure severity is correlated with severity of hypoxic-ischemic injury in abusive head trauma. Pediatr Neurol 2018;82(August 2009):29–35; doi: 10.1016/j.pediatrneurol.2017.12.003

17.

17. Ferreira‐Atuesta

, Döhler

, Erdélyi‐Canavese

, et al. Seizures after ischemic stroke: A matched multicenter study. Annals of Neurology 2021;90(5):808–820; doi: 10.1002/ana.26212

18.

18. Wang

, Vyas

, Saposnik

, et al. Incidence and management of seizures after ischemic stroke: Systematic review and meta-analysis. Neurology 2017;89(12):1220–1228; doi: 10.1212/WNL.0000000000004407

19.

19. Moran

, Lajeunesse

, Kotzur

, et al. Development of seizures following traumatic brain injury: A retrospective study. J Clin Med 2024;13(18):5399; doi: 10.3390/jcm13185399

20.

20. Yang

, Zheng

, Xu

, et al. Levetiracetam versus phenytoin for seizure prophylaxis following traumatic brain injury: A systematic review and meta-analysis. CNS Drugs 2016;30(8):677–688; doi: 10.1007/s40263-016-0365-0

21.

21. Torbic

, Forni

, Anger

, et al. Use of antiepileptics for seizure prophylaxis after traumatic brain injury. Am J Health Syst Pharm 2013;70(9):759–766; doi: 10.2146/ajhp120203

22.

22. Wang

, Chiu

, Luh

, et al. Comparative efficacy of prophylactic anticonvulsant drugs following traumatic brain injury: A systematic review and network meta-analysis of randomized controlled trials. Biagini G, ed. PLoS One 2022;17(3):e0265932; doi: 10.1371/journal.pone.0265932

23.

23. Huo

, Xu

, Li

, et al. Effectiveness of antiseizure medications therapy in preventing seizures in brain injury patients: A network meta-analysis. Front Pharmacol 2022;13:1001363; doi: 10.3389/fphar.2022.1001363

24.

24. Pandya

, Sundaram

, Baghel

. Comparison of effectiveness of brivaracetam and levetiracetam for prophylaxis of early post-traumatic seizures: A prospective comparative interventional study. Asian J Neurosurg 2024;19(4):728–734; doi: 10.1055/s-0044-1790516

25.

25. Carpay

, Aldenkamp

, Van Donselaar

. Complaints associated with the use of antiepileptic drugs: Results from a community-based study. Seizure 2005;14(3):198–206; doi: 10.1016/j.seizure.2005.01.008

26.

26. Harris

, Hateley

, Tsang

, et al. Impact of anti-epileptic drug choice on discharge in acute traumatic brain injury patients. J Neurol 2020;267(6):1774–1779; doi: 10.1007/s00415-020-09769-5

27.

27. Strzelczyk

, Schubert-Bast

. Psychobehavioural and cognitive adverse events of anti-seizure medications for the treatment of developmental and epileptic encephalopathies. CNS Drugs 2022;36(10):1079–1111; doi: 10.1007/s40263-022-00955-9

28.

28. Kellogg

, Meador

. Neurodevelopmental effects of antiepileptic drugs. Neurochem Res 2017;42(7):2065–2070; doi: 10.1007/s11064-017-2262-4

29.

29. McNamara

, Menchaca

, Abel

, et al. Effectiveness of fosphenytoin and levetiracetam to prevent posttraumatic seizures in young children with accidental or abusive traumatic brain injury. Neurocrit Care 2025;42(2):502–511; doi: 10.1007/s12028-024-02093-0

30.

30. McNamara

, Even

, Kalinowski

, et al. Multiorgan dysfunction syndrome in abusive and accidental pediatric traumatic brain injury. Neurocrit Care 2024;40(3):1099–1108; doi: 10.1007/s12028-023-01887-y

31.

31. Laing

, Gabbe

, Chen

, et al. Risk factors and prognosis of early posttraumatic seizures in moderate to severe traumatic brain injury. JAMA Neurol 2022;79(4):334–341; doi: 10.1001/jamaneurol.2021.5420

32.

32. Herman

, Abend

, Bleck

, Critical Care Continuous EEG Task Force of the American Clinical Neurophysiology Society. et al.; Consensus statement on continuous EEG in critically Ill adults and children, Part II: Personnel, technical specifications, and clinical practice. J Clin Neurophysiol 2015;32(2):96–108; doi: 10.1097/WNP.0000000000000165

33.

33. Ghossein

, Alnaji

, Webster

, et al. Continuous EEG in a pediatric intensive care unit: Adherence to monitoring criteria and barriers to adequate implementation. Neurocrit Care 2021;34(2):519–528; doi: 10.1007/s12028-020-01053-8

34.

34. Fung

, Carpenter

, Chapman

, et al. Survey of pediatric ICU EEG monitoring–reassessment after a decade. J Clin Neurophysiol 2024;41(5):458–472; doi: 10.1097/WNP.0000000000001006

35.

35. Hasbani

, Topjian

, Friess

, et al. Nonconvulsive electrographic seizures are common in children with abusive head trauma. Pediatr Crit Care Med 2013;14(7):709–715; doi: 10.1097/PCC.0b013e3182917b83

36.

36. Oh

, Olson

, Chern

, et al. Clinical characteristics and nonconvulsive seizures in young children with abusive head trauma. J Child Neurol 2019;34(12):713–719; doi: 10.1177/0883073819853973

37.

37. O’Neill

, Handler

, Tong

, et al. Incidence of seizures on continuous EEG monitoring following traumatic brain injury in children. PED 2015;16(2):167–176; doi: 10.3171/2014.12.PEDS14263

38.

38. Maguire

, Kemp

, Lumb

, et al. Estimating the probability of abusive head trauma: A pooled analysis. Pediatrics 2011;128(3):e550–e564; doi: 10.1542/peds.2010-2949

39.

39. Piteau

, Ward

MGK

, Barrowman

, et al. Clinical and radiographic characteristics associated with abusive and nonabusive head trauma: A systematic review. Pediatrics 2012;130(2):315–323; doi: 10.1542/peds.2011-1545

40.

40. Karmazyn

, Reher

, Supakul

, et al. Whole spine MRI in children with suspected child abuse. American Journal of Roentgenology 2022; doi: 10.2214/ajr.21.26674

41.

41. Narang

, Haney

, Duhaime

A-C

, AC. et al.; Abusive head trauma in infants and children: Technical report. Pediatrics 2025;155(3):e2024070457; doi: 10.1542/peds.2024-070457

42.

42. Duffy

, Squires

, Fromkin

, et al. Use of skeletal surveys to evaluate for physical abuse: Analysis of 703 consecutive skeletal surveys. Pediatrics 2011;127(1):e47–e52; doi: 10.1542/peds.2010-0298

43.

43.College of Radiology A. ACR–SPR practice parameter for skeletal surveys in children; ACR–SPR practice parameter for skeletal surveys in children. American College of Radiology 2016;1076(Revised 2008):1–9.

44.

44. Berger

, Pak

, Kolesnikova

, et al. Derivation and validation of a serum biomarker panel to identify infants with acute intracranial hemorrhage. JAMA Pediatr 2017;171(6):e170429; doi: 10.1001/jamapediatrics.2017.0429

45.

45. Hymel

, Armijo-Garcia

, Foster

, Pediatric Brain Injury Research Network (PediBIRN) Investigators. et al.; Validation of a clinical prediction rule for pediatric abusive head trauma. Pediatrics 2014;134(6):e1537–e1544; doi: 10.1542/peds.2014-1329

46.

46. Pfeiffer

, Cowley

, Kemp

, et al. Validation of the PredAHT-2 prediction tool for abusive head trauma. Emerg Med J 2020;37(3):119–126; doi: 10.1136/emermed-2019-208893

47.

47. Hazama

, Ziechmann

, Arul

, et al. The effect of keppra prophylaxis on the incidence of early onset, post-traumatic brain injury seizures. Cureus 2018;10(5):e2674–e7; doi: 10.7759/cureus.2674

48.

48. Candy

, Tsimiklis

, Poonnoose

, et al. The use of antiepileptic medication in early post traumatic seizure prophylaxis at a single institution. J Clin Neurosci 2019;69(2019):198–205; doi: 10.1016/j.jocn.2019.07.066

49.

49. DJohn

, Ibrahim

, Patel

, et al. Administration of levetiracetam in traumatic brain injury: Is it warranted? Cureus 2020;12(7):e9117; doi: 10.7759/cureus.9117

50.

50. Liou

, Chang

, Lee

, et al. Preventing epilepsy after traumatic brain injury: A propensity score analysis. J Chin Med Assoc 2020;83(10):950–955; doi: 10.1097/JCMA.0000000000000414

51.

51. Pease

, Mittal

, Merkaj

, et al. Early seizure prophylaxis in mild and moderate traumatic brain injury: A systematic review and meta-analysis. JAMA Neurol 2024;81(5):507–514; doi: 10.1001/jamaneurol.2024.0689

52.

52. Pease

, Zaher

, Lopez

, et al. Multicenter and prospective trial of anti-epileptics for early seizure prevention in mild traumatic brain injury with a positive computed tomography scan. Surg Neurol Int 2022;13:241; doi: 10.25259/SNI_38_2022

53.

53. Pingue

, Mele

, Nardone

. Post-traumatic seizures and antiepileptic therapy as predictors of the functional outcome in patients with traumatic brain injury. Sci Rep 2021;11(1):4708; doi: 10.1038/s41598-021-84203-y

54.

54. Surtees

, Kumar

, Garton

HJL

, et al. Levetiracetam prophylaxis for children admitted with traumatic brain injury. Pediatr Neurol 2022;126:114–119; doi: 10.1016/j.pediatrneurol.2021.10.009

55.

55. Fang

, Valdes

, Frontera

. Levetiracetam for seizure prophylaxis in neurocritical care: A systematic review and meta-analysis. Neurocrit Care 2022;36(1):248–258; doi: 10.1007/s12028-021-01296-z

56.

56. Romoli

, Perucca

, Sen

. Pyridoxine supplementation for levetiracetam-related neuropsychiatric adverse events: A systematic review. Epilepsy Behav 2020;103(Pt A):106861; doi: 10.1016/j.yebeh.2019.106861

57.

57.Fosphenytoin. In. LiverTox: Clinical and Research Information on Drug-Induced Liver Injury. National Institute of Diabetes and Digestive and Kidney Diseases; 2012. Available from: http://www.ncbi.nlm.nih.gov/books/NBK547879/ [Last accessed: June 30, 2025].

58.

58. Miklus

, Trout

, Even

. Levetiracetam dosing based on Glasgow Coma Scale scores in pediatric traumatic brain injury patients. Neuropediatrics 2024;55(4):224–231; doi: 10.1055/s-0044-1786796

59.

59. Kull

, Emerson

. Continuous EEG monitoring in the intensive care unit: Technical and staffing considerations. J Clin Neurophysiol 2005;22(2):107–118; doi: 10.1097/01.WNP.0000158361.24544.2D

60.

60. Lado

, Ahrens

, Riker

, National Association of Epilepsy Guidelines for Specialized Epilepsy Centers Panel. et al.; Guidelines for specialized epilepsy centers: Executive summary of the report of the national association of epilepsy centers guideline panel. Neurology 2024;102(4):e208087; doi: 10.1212/WNL.0000000000208087

61.

61. Gururangan

, Razavi

, Parvizi

. Utility of electroencephalography: Experience from a U.S. tertiary care medical center. Clin Neurophysiol 2016;127(10):3335–3340; doi: 10.1016/j.clinph.2016.08.013

62.

62. Naylor

. Treating acute seizures with benzodiazepines: Does seizure duration matter? Epileptic Disord 2014; 1(s1):S69–S83; doi: 10.1684/epd.2014.0691

63.

63. Joshi

, Rajasekaran

, Hawk

, et al. Status epilepticus: Role for etiology in determining response to benzodiazepines. Ann Neurol 2018;83(4):830–841; doi: 10.1002/ana.25213

64.

64. Mazarati

, Baldwin

, Sankar

, et al. Time-dependent decrease in the effectiveness of antiepileptic drugs during the course of self-sustaining status epilepticus. Brain Res 1998;814(1–2):179–185; doi: 10.1016/S0006-8993(98)01080-4

65.

65. Hillman

, Lehtimäki

, Peltola

, et al. Clinical significance of treatment delay in status epilepticus. Int J Emerg Med 2013;6(1):6; doi: 10.1186/1865-1380-6-6

66.

66. Gaínza-Lein

, Sánchez Fernández

, Jackson

, Pediatric Status Epilepticus Research Group . et al.; Association of time to treatment with short-term outcomes for pediatric patients with refractory convulsive status epilepticus. JAMA Neurol 2018;75(4):410–418; doi: 10.1001/jamaneurol.2017.4382

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB