Introduction

Traumatic brain injury (TBI) is an important public health issue that accounts for approximately 1.7 million emergency room visits, approximately 275,000 hospitalizations, and 52,000 deaths each year in the USA [1, 2]. Seizures are a well-known complication after TBI that can worsen the functional outcome in these patients. According to the current guidelines from the American Academy of Neurology and the Brain Trauma Foundation, seizure prophylaxis is recommended during the first week after severe TBI [2, 3].

Early post-traumatic seizures (EPTS) are defined as seizures occurring within the first week after trauma, and their incidence ranges between 2.1 and 16.9% [4]. EPTS are a strong predictor of late post-traumatic seizures and epilepsy. In fact, up to 5% of all epilepsy cases and 20% of lesional epilepsy are post-traumatic [5]. Therefore, identifying the risk factors for EPTS is very important. The aim of this study was to find the prevalence of seizures during hospitalization for TBI in a nationwide database, identify the risk factors for seizure occurrence, and assess the impact of seizure on hospital course.

Methods

Study Population

Patient data files from the National Trauma Data Bank (NTDB) between January 1, 2009, and December 31, 2010, were analyzed. NTBD is the largest and the most complete trauma database compiled by the American College of Surgeons in the USA. The database contains uniformly collected clinical, demographic, and external cause and outcome information on over three million cases from over 900 registered US trauma centers [6].

Patients aged 18 years or older with TBI were identified by the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes for TBI (800.0–801.99, 803.0–804.99, or 850.0–854.19), and similar to previous epidemiological publications on TBI [6, 7]. Variables including demographics; injury severity score abbreviated injury scale (ISSAIS); GCS score; comorbidities; type and mechanism of injury; intracranial hemorrhage; and hospital complications and outcome measurements including total hospital stay, in-hospital mortality, intensive care unit (ICU) days, ventilator days, and discharge destinations were retrieved for each patient. Seizure occurrence during hospital stay was ascertained using ICD-9-CM diagnosis code for seizure (345, 345.00-345-99, and 780.39), similar to previously published studies [8, 9].

Statistical Analysis

Seizure incidence in TBI patients was calculated, and seizure risk factors were compared with TBI patients who did not have seizures. Patient demographics, comorbid conditions, type and mechanism of injury, in-hospital complications, ICU days and length of hospital stay, ventilator days, in-hospital mortality, and discharge destination were compared between TBI patients with seizures and those without seizures. Chi-square test and t-test were used for categorical data and for continuous data, respectively, with a p value <0.05 considered statistically significant.

Multivariate linear regression was used to determine the impact of seizures on length of ICU stay, ventilator support and duration of hospital stay after adjusting for age, gender, admission GCS score, and ISSAIS. In a similar fashion, multivariate logistic regression analysis was performed to identify the impact of seizure on in-hospital mortality and discharge destination after adjusting for the confounding factors. We used the SAS 9.3 software (SAS Institute, Cary, NC) for statistical analysis.

Results

We identified a total of 360,863 patients aged 18 years or older with TBI between January 2009 and December 2010. Among these, 1559 patients (0.4%) had seizures during their hospital stay. Table 1 summarizes the demographics, basic characteristics, comorbidities, and hospital course for the seizure group (SG) and the no-seizure group (NSG).

Table 1 Demographic, baseline characteristics, comorbidities, and hospital outcome among TBI patients with and without in-hospital seizure
Full size table

The mean age of the SG was approximately 3 years higher than that of the NSG [51 (50–52) vs. 48.2 (48.1–48.3); p < 0.0001]. The rate of African-American ethnicity was 8% higher in the SG (20 vs. 12%, p < 0.0001). Similarly, white ethnicity accounted for lower percentage patients in the SG (68 vs. 70%, p = 0.004). The mean GCS score was one point lower in the SG (12 vs. 13, p = 0.45). Similar to the NSG, mild TBI (GCS score >12) occurred in more than 2/3 of subjects in the SG (71%). However, the rate of moderate TBI (GCS score 9–12) was significantly higher in the SG (8 vs. 5%, p = 0.003). The rate of sever TBI (GCS score <9) was also significantly higher among SG (21 vs. 15%), p < 0.0001). The total ISSAIS was roughly 1 points lower in the SG [13.97 (13.52–14.42) vs. 14.78 (14.74–14.82), p < 0.0001]. However, the head AIS score (it is a score of 1–6, 6 being maximum injury and non-survivable) was higher in SG [3.15 (3.08–3.22) vs. 2.87 (2.87–2.88), p ≤ 0.0001].

The mean EMS transfer time and blood pressure on hospital arrival were not different between the two groups. Subjects in the SG had significantly higher rates of smoking, obesity, hypertension, myocardial infarction, and diabetes (Table 1). Similarly, the rate of prior cerebrovascular accident (11 vs. 2%, p < 0.0001) and alcohol abuse (25 vs. 11%, p < 0.0001) was higher in the SG. A total of 13% of the SG had alcohol intoxication on hospital arrival, which was significantly lower than 24% rate among the NSG (p < 0.0001).

Intracranial hemorrhage was more common in the SG (47 vs. 36%, p < 0.0001), although no differences were seen regarding the rates of subarachnoid, epidural, or intraparenchymal hemorrhages. However, subdural hemorrhage was more common in the SG (31 vs. 21%, p < 0.0001). Pneumonia, acute respiratory distress syndrome (ARDS), acute renal failure (ARF), and pulmonary embolism were higher in the SG (Table 1). The two groups had similar rates of acute stroke during hospital stay. Notably, the rate of decompression for skull fracture was significantly higher among SG (2 vs. 1%, p = 0.013). Similarly, patients with seizure had higher rate of craniotomy during hospital stay (5 vs. 3%, p = 0.002). Of note, the SG had higher rate of blood product transfusion (15 vs. 7%, p < 0.0001) and significantly higher rate of CNS infection (0.9 vs. 0.01%, p < 0.0001).

The average length of total hospital stay was more than 3 days longer in the SG (9.6 vs. 6.3, p = 0.0002). Also, the length of ICU stay was longer among SG (7 vs. 5 day, p < 0.0001) and SG had longer average days of being on mechanical ventilation (7 vs. 6 days, p = 0.04). The SG subjects had a significantly higher rate of discharge to nursing facilities (32 vs. 25%, p < 0.0001) and a lower rate of discharge to home (60 vs. 68%, p < 0.0001). The rate of in-hospital mortality was not statistically different between two groups (8 vs. 7%, p = 0.14) (see Table 1).

Table 2 summarizes the prevalence of different mechanism of injury for both groups. While the rates of motor vehicle accidents, pedestrian injury, and firearm injury were lower in the SG, fall was the mechanism of injury in more than half of the patients in the SG (59 vs. 36%, p < 0.0001). As shown in Table 3, the rate of skull fracture was slightly lower in the SG (22 vs. 23%, p = 0.41). The rate of closed fracture was not different between the two groups. However, open skull fracture was threefold lower in the SG (0.4 vs. 1.5%, p = 0.008).

Table 2 Mechanism of injury among TBI patients with and without in-hospital seizure
Full size table
Table 3 Type of skull fracture among TBI patients with and without in-hospital seizure
Full size table

The length of hospital stay remained significantly higher in the SG after adjusting for age, ethnicity, GCS score, injury severity, comorbidities, subdural hemorrhage, CNS infection, blood transfusion, ARDS, ARF, and pneumonia [1.57 (1.42–1.69), p < 0.0001]. Similarly, the odds of discharge to a nursing facility remained significantly higher for the SG after adjusting for the aforementioned confounding factors [1.18 (1.05–1.49), p = 0.007] and these patients had lower odds of discharge to home even after adjusting for age, ethnicity, GCS score, injury severity, comorbidities, subdural hemorrhage, CNS infection, blood transfusion, ARDS, ARF, and pneumonia [0.85 (0.75–0.96), p = 0.02] (Table 4).

Table 4 Multivariate analysis* comparing hospital outcome between TBI patients with and without in-hospital seizure
Full size table

Discussion

In our analysis, seizures occurred in only 0.4% of adult patients admitted to hospitals with TBI. We found association between in-hospital seizures and age, African-American ethnicity, obesity, hypertension, diabetes, history of myocardial infarction and cerebrovascular accident, cigarette smoking, and alcohol abuse. The rate of severe TBI (GCS score ≤9) was higher among seizure group compared to non-seizure group. Moreover, seizure group had higher rates of complications including pneumonia, ARDS, acute renal failure, pulmonary embolism, and increased ICP during their hospital stay.

Early post-traumatic seizure (EPTS) is a well-known complication of TBI that can induce secondary brain injury and worsen outcome [10, 11]. It is also a well-recognized risk factor for the development of late post-traumatic seizure with a higher relative risk of developing epilepsy than the general population: 1.5 for mild TBI, 2.9 for moderate TBI, and 17 for severe TBI [10, 12, 13]. The reported incidence of EPTS varies from 1 to 26% among previous studies based on the patient age group; population selection criteria, including type of injury and severity of trauma; and use of clinical convulsive seizure as the endpoint versus EEG-based diagnosis [4, 14, 15].

We demonstrated that seizures during hospitalization for TBI are an independent predictor of longer hospital stay as well as higher probability of discharge to a nursing facility as opposed to discharge to home, even after adjusting for confounding variables such as demographics, presence of ICH, admission GCS score, injury severity, CNS infection, transfusion, and history of MI and stroke. This finding is in agreement with previous studies showing negative impact of EPTS on outcome including longer hospital stay [16]. We found no difference in the length of ICU stay between patients with seizure and those without seizure after adjusting for all confounding factors. This is similar to the results of a smaller study that demonstrated increased length of hospital stay with no increase in ICU stay in TBI patients with status epilepticus in a study of 87 patients with acute TBI requiring pediatric ICU admission [15]. In our study, the rate of in-hospital mortality was not different between SG and NSG, and this finding is similar to previously published population-based study which showed seizure does not increase in-hospital mortality rate among patients with intracerebral hemorrhage [8].

To our knowledge, our study is the first population-based study demonstrating association between African-American race and in-hospital seizure in patients with TBI. This association might be further explained by racial disparity in TBI outcome shown in previous studies [17, 18]. A recent prospective study by Ritter et al. [10] showed increased relative risk of late post-traumatic seizure among black race. Prospective studies are required to expand our understanding of impact of race on incidence of early post-traumatic seizure.

As regards to the mechanism of TBI, fall was the most common cause of TBI in patients with seizure, followed by motor vehicle accidents. While even in subjects with no seizures, falls and motor vehicle accidents were also the two most common mechanisms of TBI, and their incidence was significantly higher in the seizure group. We also found a higher rate of subdural hemorrhage in the seizure group. Remarkably, there was no difference in the rates of intraparenchymal, subarachnoid, and epidural hemorrhages between the two groups. Open skull fractures were more common in the non-seizure group. Subdural hematoma has been previously reported as strong risk factor for early seizure after TBI along with brain contusion and intraparenchymal hemorrhage [4, 12, 13].

One reason for the lower incidence of seizure in our TBI cohort may be the increased use of prophylactic anticonvulsants in our cohort which is from 2008 to 2010 compared with older reports. Also, lack of universal protocol and therefore variable utilization of continuous video-EEG monitoring (V-EEG) among participating centers in our study could have potentially affected the overall seizure detection [19]. The large nationwide database we queried does not specify whether the seizure diagnosis was based on clinical observation or EEG findings. Several studies highlighted the importance of using continuous V-EEG in the diagnosis of non-convulsive seizures [20, 21]. In the neuro-ICU, up to 34% of patients on continuous V-EEG are found to have non-convulsive seizures, of which 76% have non-convulsive status epilepticus [20, 22]. In addition, 9% of comatose patients with no history of clinical seizures are found to have non-convulsive status epilepticus when monitored by V-EEG [23]. In a prospective study of 94 patients with moderate to severe TBI who underwent V-EEG from the time of admission to the ICU up to 14 days after injury, Vespa et al. [24] reported 22% seizure incidence with more than half of the seizures being non-convulsive, which were diagnosed exclusively by EEG.

Other limitations of our study are related to utilizing the ICD-9-CM diagnostic codes [7]. Although these codes are frequently used to describe the epidemiology of TBI, different studies reported that their sensitivity and specificity range from 61 to 95% [25] [15]. In addition, TBI patients are at increased risk of developing metabolic derangement and CNS infections and are treated with a variety of medications that can be epileptogenic. It remains indeterminate whether those patients developed provoked (caused by toxins, medications or metabolic derangement), unprovoked (seizures associated with epileptic syndrome), or acute symptomatic (seizures caused by TBI, CVA, encephalitis/meningitis) seizures [26], which might have potentially affected the total number of the seizure group in this study. Also, due to the retrospective nature of our study, cautious data interpretation is required. For instance, although we identified direct association between severe TBI and higher hospital complications and early in-hospital seizures, no further conclusions regarding the cause–effect relationship between these variables can be drawn. Another potential limitation is the use of prophylactic anti-seizure medication in TBI patients, which can prevent the occurrence of acute seizures, thus affecting the number of the seizure group in our study. Due to the retrospective nature of our study, information regarding possible administration of anti-seizure medication or any sedative medications was not available.

Nevertheless, our study may have several potential clinical implications. In-hospital seizures in patients with TBI should not be underdiagnosed as they are associated with higher rates of hospital complications including pneumonia, ARDS, acute kidney injury, and increased ICP and CNS infection. Seizures also independently predict longer hospital stays and poor hospital outcome. The clinical impact of early in-hospital seizure suggests the need for more frequent use of continuous video-EEG monitoring particularly in high-risk TBI patients including patients with lower GCS score, subdural hematoma, and craniotomy. This will assist in early recognition of seizure and may result in improved outcome via implementation of appropriate management.