Introduction

Traumatic brain injury (TBI) is a major burden to public health, with its incidence worldwide estimated to be in excess of 10 million cases a year [1]. Pediatric TBI is estimated to occur in 70 per 100,000 children, and its disease burden is disproportionately high due to the age of these patients and the longer time they have in which post-trauma burdens can manifest [2, 3]. Although pediatric hospitalization and mortality rates for mild traumatic brain injuries have decreased over the past decades, rates for moderate and severe traumatic brain injuries are relatively unchanged [4]. As such, there remains a need to profile pediatric TBIs to alleviate disease burden on both hospitals and patients moving forward.

Clinically, pediatric TBI is a very heterogeneous diagnosis. Depending on patient characteristics and injury etiology, inpatient admissions for head injury vary between short observations and rapid discharge home, to longer hospitalizations following surgical intervention or intensive care. A better understanding of these presentations will enhance our current understanding of what constitutes optimal management and resource allocation to further optimize hospital and patient outcomes [5]. Correspondingly, the objectives of this study were to assess the characteristics, etiology, and clinical inpatient outcomes of pediatric patients with TBI, focusing on identifying predictors of injury severity and short hospitalization to better inform policy.

Methods

Study design

We conducted a retrospective cohort analysis of children with TBI by interrogating the Kids’ Inpatient Database (KID), the largest publicly available all-payer pediatric inpatient care database in the USA. We examined the characteristics of head injury and TBI among children ≤ 20 years using a population-based database published every 3 years, encompassing a large and representative cohort of pediatric inpatient hospitalizations. Four years of KID data (2006, 2009, 2012, and 2016) spanning a decade were analyzed. Diagnosis of head injury and associated comorbidities and procedures using the ICD 9CM/PCS (2006, 2009, 2012) or ICD 10 CM/PCS (2016) were analyzed [6]. ICD 10 codes for pediatric head injury included open wounds of the head and skull fractures (S01.7–9, S02.0, 0.1, 0.3, 0.7–9), injuries of optic nerve and pathways (S04.0), intracranial injuries (S06), and multiple other head injuries (S07, S09.7–9), alongside their ICD 9 counterparts were included (Supplemental Table 1). Although the ICD definition of TBI and head injury varies and may include other codes, the above codes were chosen to provide accuracy and consistency across other pediatric head injury and TBI studies [7]. Codes characterizing sequelae of injury were not included as we select for initial inpatient visits for TBI. Hospital births were removed to exclude birth-related injuries. The severity of TBI was calculated by the Injury Severity Score (ISS), a common and validated scale for assessment of traumatic injuries, using the Stata program International Categorization of Disease Programs for Injury Categorization (ICDPIC) [8].

Table 1 Demographic characteristics of all included cases. Unless stated otherwise, continuous data presented as mean ± standard error, categorical data presented as percentage of total
Full size table

Statistical analysis

Patient characteristics were noted with descriptive statistics, continuous variables with means ± standard error (SE), and categorical variables with frequencies and percentages. Primary summary statistic was odds ratio (OR). Predictors of increased injury severity, characterized by Injury Severity Score (ISS), were analyzed using multivariate linear regression models. A group of patients with short hospitalization (≤ 1 day hospitalization) were identified and various factors associated with their inpatient stay were analyzed. For both linear and logistic regression models, statistical significance of P < 0.20 was included on the final multivariate model and statistical associations were considered significant for P < 0.05. Data were analyzed using Stata 13 (College Station, TX: StataCorp LP).

Results

TBI cohort description

There were a total of 220,777 pediatric TBI cases across the entire dataset of 12,809,624 admissions, equaling an incidence of 1.7% (Table 1). The majority of cases are boys (66%), with an average age of 11.5 ± 0.02 years. Admissions over the weekend constituted of 34% of presentations, with 17% transferred to the hospital. The most and least common seasons for admission were summer (28%) and winter (21%) respectively. In terms of pre-existing conditions prior to injury, 6% reported chronic alcohol use, 5% chronic drug use, and 4% neurological issues. In terms of the patient, the most common financial allowance was private insurance (47%), and most common household income was within the first quartile (30%). Hospitals were most commonly private, not for profit (67%) of large size (81%) and predominantly teaching hospitals (81%) in nature. For outcomes, mean LOS was 5.0 ± 0.03 days, with 25% discharged within 1 day, and 83% routinely discharged home. In-hospital mortality occurred in 4% of cases.

Injury etiology

Pediatric TBIs were dominated by motor vehicle (MV) accidents (29%), followed by falls (19%), assaults (9%), and non-MV traffic injuries (5%) (Table 2). MV injuries tended to present in older patients (mean age 14.8 years), whereas falls presented in younger patients (mean age 7.4 years). Compared to all presentations, proportion of male patients was higher in assault injuries (79% vs 67%). When compared to fall and assault injuries, MV and non-MV traffic injuries trended towards greater proportion of weekend admissions, higher ISS, greater representation of profound injuries, longer LOS, lower routine discharges to home, and higher in-hospital mortality rates. The longest LOS by etiology was following MVT in a median 6 days (range 0–166 days) whereas falls had the shorted LOS in a median of 2 days (range 0–61 days) (Fig. 1).

Table 2 Descriptive characteristics of four most common external etiologies of traumatic brain injury (TBI). Unless otherwise stated, continuous data presented as mean ± standard error, categorical data presented as percentage of total
Full size table
Fig. 1
figure 1

Column graph of mean length of stay by etiology of traumatic brain injury, upper bar represents maximum range limit. MVT, motor vehicle traffic

Full size image

Predictors of injury severity

More severe TBI presentations were significantly and independently associated with older age, weekend admissions, hospital transfers, in patients with chronic conditions, and neurological issues (all P < 0.01) (Table 3). Chronic alcohol use was significantly associated with decreased severity of TBI (P < 0.01). Hospitals that were government-funded, larger in size, and holding teaching status were also significantly and independently associated with more severe TBI presentations (all P < 0.01). Although TBI in children tended to occur most frequently in the summer, there were no differences in the severity of trauma between seasons (all P > 0.05).

Table 3 Uni- and multivariate analyses evaluating associations with increased injury severity score (ISS), with coefficient as summary statistic
Full size table

Predictors of short hospitalization after TBI

Within all pediatric TBI presentations, 38% were discharged within 1 day of admission. Younger age, male gender, less chronic conditions, fall and assault etiologies, with milder injury severity, and private insurance all significantly and independently predicted greater likelihood of short hospitalization (all P < 0.01) (Table 4). Presentations to teaching hospitals and the requirement of neurosurgical intervention both significantly and independently predicted lower likelihood of 1-day stay (all P < 0.01). The most common neurosurgical procedure performed involved elevation of skull fracture fragments/extirpation of matter from the epidural space, accounting for 40% of those procedures.

Table 4 Uni- and multivariate analyses evaluating associations with short stay hospitalization (length of stay ≤ 1 day) versus non-short study hospitalizations, with odds ratio (OR) as summary statistic. Unless otherwise stated, continuous data presented as mean ± standard error, categorical data presented as percentage of total. ISS, injury severity score
Full size table

Discussion

Understanding the nature of pediatric TBI is important from a management perspective to ensure adequate neurosurgical resources are appropriately allocated. Our study provides a contemporary survey on the current pediatric TBI climate. We have identified a number of clinical associations both injury severity and short hospitalization, which may be used in the future to predict appropriate level of care and expectations to pediatric TBI presentations.

The finding that older pediatric males are more likely to present with TBI to large, urban teaching hospitals reaffirms the findings of broader pediatric emergency studies. When Dewan et al. [9] surveyed the contemporary literature, they found trends that male children were more susceptible to TBI than females, as well as identifying that within the USA, children aged 14 years and older had greater odds of sustaining a non-abuse TBI, findings which have been reproduced by analysis of various national databases and registries from around the world [3, 10, 11]. Presentation to large, urban teaching hospitals has been noted previously [12], with the greater populations in urban areas compared to rural areas a likely mechanism of this numerical disparity [13].

In terms of injury etiology, our study found that in the USA, the most common were MVTs, falls, assaults, and non-MVT traffic accidents, findings that have been reported elsewhere [14, 15]. However, our study was able to discern differences in clinical course which have not been previously discussed. Of those etiologies, MVT and other traffic injuries were associated with the higher severity, higher mortality rates, and longer LOS, whereas assault TBIs were more likely sustained in male patients than female patients. Furthermore, MVT was associated with the lowest rate of disposition to home, and thus the highest rate of disposition to rehabilitation. Knowledge of which etiologies predispose patients to rehabilitation the most can assist hospital administrators in initiating pre-emptively rehabilitation referrals requiring insurance validation. Insurance itself is associated with short hospitalization per our study. By minimizing the time from admission to financial approval, active initiative can reduce the length of stay in the hospital should these patients be transferred to rehabilitation earlier. This is important because the adult literature suggests that the earlier the rehabilitation, the more improved the final prognosis following TBI [16].

Our study also provides insight into associations of TBI severity in children. Of note was the observation that increased chronic conditions, such as alcohol consumption and neurological issues, increased risk of severity. Therefore, a thorough history is crucial in working up pediatric TBI at initial presentation. In the USA, the legal drinking age is 21 years, and therefore, this correlation reveals that alcohol testing should be considered when evaluating older pediatric TBI presentations, and that counselling services should feature in the management of injuries that are positive for alcohol consumption. Inversely, severity is less likely to correlate to substance use Similarly, knowledge of neurological issues should be addressed in post-hospitalization management as well to ensure that all measures are taken to minimize further injury risk to the patient.

In all, approximately one-third of pediatric TBI patients were discharged following a short hospitalization stay of 1 day in the hospital. These children tend to be younger children with low-impact injury mechanisms and mild injury as defined by ISS. This would favor incorporating the ISS as a metric in triaging pediatric TBI when mild injury is suspected in the setting of a normal and non-suspicious neurologic history and exam. It is worth noting that incorporation of ISS score into adult TBI algorithms for triage already has been reported [17]. Optimal triaging may in the future better define transfer thresholds based on these parameters to avoid overburdening larger hospital systems with these short stay presentations that can be adequately managed at smaller hospitals [5]. This can include strict criteria for radiographic findings, such as the absence of intracranial hemorrhage, and clinical findings, such as Glasgow Coma Scores of 15, which all implicate a less care-intensive TBI.

Despite a particular degree of intuition, particular elements of our findings are worth highlighting. Firstly, it is interesting that seasonality did not predict trauma severity. Timing has not been largely reported in pediatric TBI specifically, with the only national epidemiological study [18] of all pediatric patients from South Africa, and an institutional experience [19] with infants in France, both reporting that most injuries they recorded occurred on weekends. Similarly with respect to seasons, few studies have focus on pediatric TBI studies although those that have report that spring and summer months are seasons with increased presentations compared to winter both in the USA [20] and elsewhere [21]. Otherwise, despite increased presentations during particular months, the implication of our study is that the proportion of severe traumas remains steady throughout the year.

Another finding worth mentioning is the most common indications for neurosurgical intervention were extracranial in nature, rather than intracranial. In the setting of head trauma, this ratio in pediatric TBI would appear to be inverse to that of adult TBI [22], highlighting the need for separate evaluation algorithms for pediatric TBI than adult counterparts. Yet if there is ever concern about the surgical candidacy of a pediatric TBI presentation, transfers should be initiated to larger tertiary centers with neurosurgical services should the initial institution not be able to provide this care.

There are limitations to this study inherent in its design. Firstly, it is a retrospective study of registry data which cannot be expanded upon. Although this limits the breadth of inferences possible, we are able to hypothesize parameters that can start to assist in triaging pediatric TBI to the appropriate level of care. Ultimately, more prospective granular data is needed to validate these suspicions, and furthermore build prediction models for when transfer to higher level centers is needed. Data on transfer to high-level centers of care is not available in the KID dataset, and would require multi-institutional collaborations to establish this very important element in pediatric TBI management.

Secondly, the generalizability of our results may not extend to all settings. There is a growing body of evidence [9, 23] to suggest that mechanisms (and therefore outcomes) of pediatric TBI vary between geographic regions of the world, meaning that our results may be most applicable to comparable Western countries. Finally, utilization of an inpatient database means that we are not able to establish long-term post-hospitalization outcomes in these patients such as impacts on domains of learning, emotional awareness, and social functioning [24]. It is possible that these outcomes will also guide appropriate clinical and social follow-up for these pediatric TBI patients, and should be integrated concurrently with inpatient clinical course to optimize patient care.

Conclusion

Pediatric TBI is a potentially life-threatening presentation which mandates that hospital resources be allocated appropriately. In our study, we have identified multiple clinical parameters that profile pediatric TBI, as well as present a series of contemporary clinical associations for both injury severity and short hospitalization of 1-day inpatient stay. Better understanding of how pediatric TBI patients present will better our ability to triage severity in the future.