Traumatic brain injury (TBI) is a leading cause of morbidity

Transcription

1 Assessment of Recovery Following Pediatric Traumatic Brain Injury Julia C. Slovis, MD 1 ; Nachi Gupta, MD, PhD 2 ; Natasha Y. Li, MD 1,3 ; Steven G. Kernie, MD 1 ; Darryl K. Miles, MD 4 Objectives: We analyzed a prospective database of pediatric traumatic brain injury patients to identify predictors of outcome and describe the change in function over time. We hypothesized that neurologic status at hospital discharge would not reflect the longterm neurologic recovery state. Design: This is a descriptive cohort analysis of a single-center prospective database of pediatric traumatic brain injury patients from 2001 to Functional outcome was assessed at hospital discharge, and the Glasgow Outcome Scale Extended Pediatrics or Glasgow Outcome Scale was assessed on average at 15.8 months after injury. Setting: Children s Medical Center Dallas, a single-center PICU and Level 1 Trauma Center. Patients: Patients, 0 17 years old, with complicated-mild/moderate or severe accidental traumatic brain injury. Measurements and Main Results: Dichotomized long-term outcome was favorable in 217 of 258 patients (84%), 80 of 82 patients (98%) with complicated-mild/moderate injury and 133 of 172 severe patients (77%). In the bivariate analysis, younger age, motor vehicle collision as a mechanism of injury, intracranial pressure monitor placement, cardiopulmonary resuscitation at scene or emergency department, increased hospital length of stay, increased ventilator days (all with p < 0.01) and occurrence of seizures (p = 0.03) were significantly associated with an unfavorable outcome. In multiple regression analysis, younger age (p = 0.03), motor vehicle collision (p = 0.01), cardiopulmonary 1 Department of Pediatrics, Columbia University Medical Center, New York- Presbyterian Morgan Stanley Children s Hospital, New York, NY. 2 Department of Emergency Medicine, Icahn School of Medicine at Mount Sinai, New York, NY. 3 Department of Pediatrics, Loma Linda University Medical Center, Loma Linda, CA. 4 Department of Pediatrics, The University of Texas Southwestern Medical Center, Children s Medical Center Dallas, Dallas, TX. Supported, in part, by Perot Family Center for Brain and Nerve Injuries at Children s Medical Center Dallas. The authors have disclosed that they do not have any potential conflicts of interest. For information regarding this article, darryl.miles@utsouthwestern.edu Copyright 2018 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies DOI: /PCC resuscitation (p < 0.01), and ventilator days (p < 0.01) remained significant. Remarkably, 28 of 60 children (47%) with an unfavorable Glasgow Outcome Scale at hospital discharge improved to a favorable outcome. In severe patients with an unfavorable outcome at hospital discharge, younger age was identified as a risk factor for remaining in an unfavorable condition (p = 0.1). Conclusions: Despite a neurologic status at hospital discharge, many children after traumatic brain injury will significantly improve at long-term assessment. The factors most associated with outcomes were age, cardiopulmonary resuscitation, motor vehicle collision, intracranial pressure placement, days on a ventilator, hospital length of stay, and seizures. The factor most associated with improvement from an unfavorable neurologic status at discharge was being older. (Pediatr Crit Care Med 2018; XX:00 00) Key Words: critical care outcomes; Glasgow Outcome Scale; pediatric intensive care units; post-traumatic vegetative state; traumatic brain injury Traumatic brain injury (TBI) is a leading cause of morbidity and death in the pediatric population (1), but it remains difficult to predict long-term outcome. The prognosis is overall favorable for mild TBI (1), but severe TBI often results in considerable long-term neurologic disabilities. After recovery from the initial injury, head-injured children may exhibit deficits in executive function, memory, cognition, motor function, seizures, psychiatric disorders, and emotional and behavioral problems (2 4). Clinicians and families are often faced with the burden of this uncertainty in weighing the possible long-term care outcomes. Assessing the functional impairments after TBI in children is challenging considering the broad range of assessment possibilities, developmental stages, and the heterogeneity of brain injuries. Studies conflict on whether the immature and developing brain is more vulnerable or resilient following traumatic injury (5), but it has been hypothesized that arrested neural development after injury at a younger age confers a worse prognosis (4, 6). Illuminating predictive variables for recovery following TBI can help identify potentially treatable factors of secondary brain injury and may assist in a more reliable counseling discussion regarding long-term expectations. Many Pediatric Critical Care Medicine 1

2 Slovis et al ethical concerns may arise during the acute injury phase from a patient s potential loss of ability for neurologic recovery, including the possibility of withdrawing life supportive care or the decision to perform technology-dependence procedures such as a tracheostomy and chronic ventilation. A crucial question for both families and clinicians is, what is the likely recovery potential of neurologic function over time? Outcome data from long-term assessments remote from the injury are essential when forming these decisions (7). Neurologic recovery is known to occur in TBI over time and further understanding the recovery potential is both clinically and socially relevant to help guide treatment and expectations. The objective of the current study was to examine data from a prospective database of a large cohort of pediatric patients, who suffered complicated-mild (cmild)/moderate or severe accidental TBI, in order to determine prognostic injury and hospital factors of longterm outcome and to assess the change in neurologic function from hospital discharge to long-term follow-up assessments. Functional outcome was assessed at hospital discharge and again between 6 and 18 months after TBI. Logistic regression bivariate and multivariate statistical models were used to evaluate pre-hospital, injury, and hospital course variables with long-term functional outcome and to compare outcomes at discharge to recovery outcome groups. Few large cohort studies have assessed early prognostic variables after pediatric TBI measured against outcomes at 1-year from injury (7) or analyzed the change in function over time compared with outcomes at hospital discharge excluding abusive head injury (1, 6 9). The Glasgow Outcome Scale (GOS) is a five-point scale developed to categorize outcomes after adult and pediatric TBI, representing a progressive scale from death, vegetative state, independence in the home, return to work, and the re-establishment of normal social relationships. The Glasgow Outcome Scale Extended Pediatrics (GOSEP) scale, adapted from the adult GOS Extended scale, was also used for long-tem functional outcome analysis. This outcome measure validated for use in children is an eight-point scale adding increased discrimination of functional disability scores within the normal, mild, moderate, and severe categories (10). We hypothesized that outcomes would significantly improve at long-term follow-up compared with hospital discharge and that children after TBI would move meaningfully in outcome groups in a favorable direction. We proposed that patients improving from an unfavorable outcome at discharge to favorable outcome category at long-term follow up would represent a meaningful change, but given the broad scope of disabilities encompassed within the GOS, even a one-point improvement may have a significant impact on patients and their families. MATERIALS AND METHODS Patient Data Collection A prospectively collected dataset from the Brain and Nerve Injury Center Repository and Database (BNIRD) at Children s Medical Center Dallas of 258 recruited patients, 0 17 years old, between 2001 and 2012 (95% were < 13 yr) was analyzed. Outcomes and clinical study variables were abstracted from a separate prospective database for patients prior to the formation of the BNIRD in 2005, using a similar protocol. Long-term functional outcome was assessed by study personnel using the GOS from 2001 to 2004, and the GOSEP from 2005 to All subjects were enrolled after written parental consent with approval from the UT Southwestern (UTSW) Medical Center Institutional Review Board and materials data transfer agreement between UTSW and Columbia University Medical Center. TBI was secondary to blunt force trauma from accidental mechanism, and the postresuscitation emergency department (ED) GCS assessed neurologic injury severity. For statistical analysis, the GCS was dichotomized into severe (GCS, 3 8) and cmild/moderate (GCS, 9 15) injury (11). Complicated mild injury was defined as a GCS of with evidence of intracranial injury on head CT, and moderate injury was defined as a GCS of These two categories were combined to dichotomize severe versus cmild/moderate. Children were excluded if the TBI was penetrating or abusive or if the study team thought the child s injuries were likely nonsurvivable. Research staff collected deidentified demographic, clinical, and radiographic data and retrospectively supplemented the database with medical history, diagnostic tests, operative reports, neuroimaging findings, and discharge GOS. A trained member of the study research team prospectively assessed and collected the discharge GOS prior to hospital discharge. Variables and interventions were assembled from the scene (if available), from the ED, and from the hospital course. All analyses were based only on available data, and the data that is missing was considered to be missing at random. Patients received physical and cognitive evaluations prior to hospital discharge by a member of physical medicine and rehabilitation team. The discharge disposition was based on these assessments along with the primary trauma or neurosurgical team s recommendation. Data were collected using similar methods and personnel throughout the 12-year period. In 2003, the Brain Trauma Foundation published a set of guidelines for pediatric TBI (12). After publication of these guidelines, patients were treated according to these guidelines, but apart from these standards, care was not protocol-driven. Neurologic Outcome Assessments Long-term neurologic assessment outcomes were collected prospectively between 6 and 18 months after injury by either phone call or in-office assessment and entered into the database. For patients with more than one follow-up neurologic assessment between 6 and 18 months, the assessment closest to the 12-month period was used for the primary outcome measure. For simplicity, in the text we use more than 6 months as the time point for long-term recovery to represent the followup time frame. The eight-point GOSEP and five-point GOS were dichotomized into favorable and unfavorable outcome categories for statistical analysis. A GOS of 5 equal to normal/ mild disability or 4 equal to moderate disability or a GOSEP score of 1 equal to normal, 2 equal to mild disability, 3 equal 2 XXX 2018 Volume XX Number XXX

3 Neurocritical Care to upper moderate disability, or 4 equal to lower moderate disability was classified as a favorable outcome. A GOS of 3 equal to severe disability, 2 equal to vegetative state, or 1 equal to death or a GOSEP score of 5 equal to upper severe disability, 6 equal to lower severe disability, 7 equal to vegetative state, or 8 equal to death was classified as an unfavorable outcome. Of the 258 outcomes scores at long-term follow-up ( 6 mo), 137 were GOSEP scores, obtained after 2005 through face-toface or phone interviews conducted by a neuropsychologist or trained neuropsychology technician. GOSEP was not used prior to 2005 because the scale was not available prior to that year. Statistical Analysis Descriptive statistics are presented as means, ranges, and percentages. Logistic regressions were done with the outcome of interest being a binary representation of the long-term outcome at greater than or equal to 6 months as favorable (given a value of 1) and unfavorable (given a value of 0). Only patients with a severe TBI (ED GCS, 3 8) were included in the statistical model regressions. A bivariate analysis was done first to assess the relationship of each predictive variable independently on the long-term outcome. A multiple logistic regression was then done to assess the relationship of variables jointly on the longterm outcome. Continuous variables in the regression were age, ventilator days, and hospital LOS; all other variables were binary. To assess which factors may be associated with change in long-term function, measured by moving from an unfavorable function group at hospital discharge to favorable outcome category at long-term follow-up, a separate analysis was done using only the severe patients who had an unfavorable hospital discharge status. Bivariate and multiple logistic regressions were done again on just this subset of patients with the outcome of interest being the same binary representation of longterm outcome. Note that some of the variables could not be included in these regressions because the data for those variables was too limited when looking at this subset of patients. A p value of less than 0.05 was used for statistical significance. All analysis was done in the R Language and Environment for Statistical Computing version RESULTS Patient Demographics and Injury Characteristics Descriptive statistics for cmild/moderate and severe TBI patients are listed in Table 1. The mean age at injury was 79 months, and 66% of patients were males. The average GCS in the ED was 13 for the cmild/moderate cohort and 4 for the severe TBI patients. Severe TBI occurred in 172 of 254 (68%), and cmild/moderate injury occurred in 82 of 254 (32%) of subjects. There were eight deaths among study patients (3% mortality), and all were in the severe TBI group. These patients were included in all analyses. The primary mechanisms of injury were motor vehicle collisions (MVCs) (53%), motor pedestrian collisions (21%), patient falls (21%), and bicycle accidents (7%). Seizures (clinical or electrographic) occurred in the field or hospital in 41 of 258 (16%), and the average time on a ventilator was 5 days. A favorable outcome at hospital discharge occurred in 164 of 224 (73%) for all patients, 69 of 72 (96%) were cmild/moderate injury, and 91 of 148 (61%) had severe injury. A long-term outcome ( 6 mo) was favorable in 217 of 258 (84%) of all patients, 80 of 82 (98%) were cmild/ moderate injury and 133 of 172 (77%) with severe TBI. Four patients did not have a GCS recorded, which were included in the all category but not in subgroup analysis. Injury and Hospital Variable Associations With Long- Term Outcomes Bivariate and multiple logistic regression analysis of patient, injury, and hospital variables associated with long-term outcomes are presented in Table 2. We included only patients with severe TBI (n = 142) with complete data points in this analysis as only 2 of 82 patients (1%) in the cmild/moderate category sustained an unfavorable long-term outcome. Of note, all patients who sustained falls 55 of 55 (100%) had a favorable outcome. In the bivariate analysis, younger age, MVC as a mechanism of injury, intracranial pressure (ICP) monitor placement, cardiopulmonary resuscitation (CPR) at scene or ED, increased hospital LOS, increased ventilator days (all with p < 0.01) and occurrence of seizures (p = 0.03) were significantly associated with an unfavorable outcome. In multiple regression analysis, younger age (p = 0.03), MVC (p = 0.01), CPR (p < 0.01), and ventilator days (p < 0.01) remained significant. ICP monitor placement was not statistically significant (p = 0.19) in the multiple regression. For severe patients with no ICP monitor placed, there were 74 patients, 55 of 62 (89%) had a favorable outcome at hospital discharge (12 missing values) and 69 of 74 (93%) had favorable long-term outcome. We did not find any significant associations with hypotension, hypoxia, fever, or pupillary examination with long-term outcome in severe TBI, but the small numbers in these categories may have resulted in insufficient power to detect an effect. Recovery After Hospital Discharge: Comparison of Short- and Long-Term Functional Outcomes The majority of children with severe and cmild/moderate head injury had a favorable discharge and long-term outcome, but important distinctions were apparent when examining the two injury groups and examining the recovery progression over time. In general, patients who were discharged from the hospital with a favorable neurologic function score stayed in the favorable outcome category at long-term assessment, 86 of 91 (95%) for the severe injury group and 68 of 69 (99%) for the cmild/moderate group. Five patients in the severe TBI group and one patient in the cmild/moderate group declined from a favorable to an unfavorable outcome at long-term follow-up assessment. Of the 60 patients with an unfavorable discharge, neurologic function (57 severe and three cmild/moderate) at hospital discharge, roughly half, 28 of 60 (47%) went on to improve to a favorable outcome category with recovery over time after discharge (Table 3). In eight patients with a vegetative Pediatric Critical Care Medicine 3

4 Slovis et al TABLE 1. Descriptive Statistics for Complicated Mild/Moderate and Severe Traumatic Brain Injury Patients Demographic, Injury, and Hospital Variables Complicated mild/moderate Traumatic Brain Injury (n = 82) Severe (n = 172) All (n = 258) Mean age (mo) (range) 70 (0 178) 82 (3 205) 79 (0 205) Male gender, n (%) 53 (65) 113 (66) 169 (66) Mechanism of injury, n (%) Motor vehicle accident 23 (28) 74 (43) 97 (38) Patient fall 27 (33) 27 (16) 55 (21) Motor pedestrian collision 14 (17) 36 (21) 53 (21) Bicycle 9 (11) 9 (5) 18 (7) Falling object 4 (5) 10 (6) 14 (5) All terrain vehicle or motorbike 3 (4) 8 (5) 12 (5) Other mechanism 2 (2) 8 (5) 10 (4) Intracranial pressure monitor placed, n (%) 15 (18) 98 (57) 114 (44) Cardiopulmonary resuscitation (at ED or scene), n (%) 1 (1) 12 (7) 13 (5) Loss of consciousness, n (%) 26 (32) 117 (68) 145 (56) Hypotension (at ED or scene) a, n (%) 1/82 (1) 13/171 (8) 15/257 (6) Hypothermic (at ED) a, n (%) 17/76 (22) 44/153 (29) 62/233 (27) Febrile (at ED) a, n (%) 4/76 (5) 11/153 (7) 16/233 (7) Hypoxia a, n (%) 6/59 (10) 20/157 (13) 26/219 (12) Intubated by emergency medical services a, n (%) 3/73 (4) 86/170 (51) 92/247 (37) Intubated by ED or outside hospital a, n (%) 26/73 (36) 81/170 (48) 107/247 (43) Pupillary examination b, n (%) Bilaterally equal and reactive 31/32 (97) 59/88 (67) 92/122 (75) Bilaterally nonreactive 0/32 (0) 15/88 (17) 15/122 (12) Unilaterally fixed/dilated 0/32 (0) 9/88 (10) 9/122 (7) Bilaterally asymmetric/reactive 1/32 (3) 5/88 (6) 6/122 (5) Mean Glasgow Coma Score in ED a, (sd) 13 (2) 4 (2) 7 (4) Hospital LOS days, mean (range) 7 (0 39) 14 (0 47) 12 (0 47) Ventilator days, mean (range) 2 (0 20) 7 (0 36) 5 (0 36) PICU LOS days, mean (range) 4 (0 19) 10 (0 38) 8 (0 38) Seizures in the hospital 13/82 (16) 28 (16) 41 (16) Discharge disposition a, n (%) Inpatient rehabilitation 10/82 (12) 89/163 (55) 98/243 (40) Home 72/82 (88) 66/163 (40) 137/243 (56) Morgue 0 8/163 (5) 8/243 (3) Favorable Glasgow Outcome Score at hospital discharge a 69/72 (96) 91/148 (61) 164/224 (73) Favorable long-term outcome 6 mo 80/82 (98) 133/172 (77) 217/258 (84) ED = emergency department, LOS = length of stay. a For these categories, there were missing data points. The number of events/total data points available and the percentage of the ratio are presented in each cell. b Pupillary data were not available before Four patients did not have a Glasgow Coma Score reported. Hypothermia = temperature 36.0 C; hypotension = systolic blood pressure < 70 mm Hg (+ age [yr] 2); hypoxia = systemic oxygen saturation 88%; febrile = temperature 38.0 C. Favorable long-term outcome 6 months = (Glasgow Outcome Score of 5 = normal/mild disability or 4 = moderate disability) or (Glasgow Outcome Scale Extended Pediatrics score of 1 = normal, 2 = mild disability, 3 = upper moderate disability, or 4 = lower moderate disability). 4 XXX 2018 Volume XX Number XXX

5 Neurocritical Care TABLE 2. Bivariate and Multiple Logistic Regressions of Injury Variables in Severe Traumatic Brain Injury Patients With a Reported Long-Term Functional Outcome (n = 142) Demographic, Injury, and Hospital Variables Bivariate Multiple β a p β a p Age 0.01 < Male gender Mechanism of injury Motor vehicle collision 0.21 < Motor pedestrian collision Intracranial pressure monitor placed Cardiopulmonary resuscitation at ED or scene < < < 0.01 Loss of consciousness Hypotension at ED or scene Hypothermic in ED Febrile in ED Hypoxia prior to hospital admission Pupillary examination Bilaterally equal and reactive Bilaterally nonreactive Unilaterally fixed/dilated Bilaterally asymmetric/ reactive Hospital length of stay (d) 0.01 < Ventilator days 0.02 < < 0.01 Seizures ED = emergency department. a β is the coefficient of the variable from the logistic regression. Hypothermia = temperature 36.0 C; hypotension = systolic blood pressure < 70 mm Hg (+ age [yr] 2); hypoxia = systemic oxygen saturation 88%; febrile = temperature 38.0 C. Long-term 6-month functional outcome was dichotomized into favorable good or unfavorable poor categories by Glasgow Outcome Score or Glasgow Outcome Scale Extended Pediatrics. state at discharge who had follow-up data, seven ultimately regained consciousness (88%). Two of the eight (25%) eventually developed favorable neurologic outcomes as defined by a favorable GOS. We next examined whether we could identify any patient, injury, or hospital variables that would potentially be associated with the observed recovery progression from an unfavorable status at hospital discharge to favorable outcome at long-term follow-up in the severe TBI patients. When examining only the patients with a poor outcome at hospital discharge (n = 57), some variables were unable to be included in the regression analysis due to missing data points or insufficient factor variability. In 57 children with severe TBI and an unfavorable status at hospital discharge bivariate (p = 0.01) and multiple logistic regression (p = 0.05) found a significant relationship where younger age was associated with the probability of continuing to have an unfavorable outcome if discharged with an unfavorable status. No other factors were significant in this analysis (Table 4). This suggests that intrinsic mechanisms specific to the younger age child after TBI may in part lead to a reduced recovery potential contributing to their increased risk for remaining in an unfavorable long-term outcome after hospital discharge. DISCUSSION TBI in childhood has an overall favorable prognosis, yet considerable neurologic disabilities still occur in a significant percentage of survivors. In a large cohort of pediatric headinjured patients, we identify risk factors predicting neurologic outcome after childhood TBI and examine the change in neurologic recovery from hospital discharge to long-term neurologic assessment between 6 and 18 months after injury. The role of mechanism of injury in outcome prognostication, excluding nonaccidental trauma, has not been fully elucidated, and the present study is one of a few to assess only accidental injury in children (7, 13). Prior studies suggest that injury mechanism is significantly related to outcome; however, these studies included nonaccidental trauma (1, 6, 7, 9, 14) gun shot wounds (15) or compare specific injury mechanisms (e.g., closed vs penetrating) (16). We found a significant favorable outcome association with falls as a mechanism of injury, of the 55 patients with falls, all had a favorable long-term outcome, despite 27 of 55 (49%) having an initial GCS of 3 8. This is consistent with prior reports showing that patients with falls frequently have a favorable outcome despite having an initial GCS of less than or equal to 8 on presentation (15). In contrast, we found MVCs to be associated with a higher probability for an unfavorable outcome, showing MVCs as a particularly hazardous insult to neural tissue (17). In falls, blunt force impact is the predominant force transmitted to the brain, whereas significant acceleration deceleration and angular forces can be experienced in high speed MVCs, leading to more diffuse injury and potentially impairing recovery. We also observed a significant association using multivariate logistic regressions with CPR at the scene or ED, ventilator days, and younger age with outcome. We were surprised to not find a significant association with variables previously shown to be associated with outcome in children such as the pupillary examination, hypoxia, and hypotension (7, 9, 14, 18, 19). Interestingly, a small number of the severe TBI cohort presented with hypotension (8%) or hypoxia (13%). One explanation for the lack of significance for these Pediatric Critical Care Medicine 5

6 Slovis et al TABLE 3. Change in Neurologic Outcome From Hospital Discharge to 6 Months in Traumatic Brain Injury Patients Timing of Outcome Assessment Outcome Category at 6 Mo Severe TBI (GCS 3-8) Complicated Mild/Moderate TBI (GCS 9-15) Unfavorable GOS Favorable GOS Unfavorable GOS Favorable GOS Hospital discharge, n (%) 57/148 (39) 91/148 (61) 3/72 (4) 69/72 (96) > 6 mo, n (%) Unfavorable Favorable 31/57 (54) 26/57 (46) 5/91 (5) 86/91 (95) 1/3 (33) 2/3 (67) 1/69 (1) 68/69 (99) GCS = Glasgow Coma Score, GOS = Glasgow Outcome Score, TBI = traumatic brain injury. This table includes patients with a GOS at hospital discharge and a long-term outcome score. Note that the number and percentages of patients in the outcome at 6 months rows are represented with the total number of outcomes in the denominator. TABLE 4. Bivariate and Multiple Logistic Regressions of Variables Associated With Long-Term Outcome in Severe Traumatic Brain Injury Patients With Unfavorable Glasgow Outcome Score at Hospital Discharge (n = 57) Demographic, Injury, and Hospital Variables Bivariate Multiple β a p β a p Age Gender Mechanism of injury Motor vehicle collision Motor pedestrian collision Loss of consciousness Hypothermic in emergency department Hypoxia prior to hospital admission Pupillary examination Bilaterally equal and reactive Bilaterally nonreactive Hospital length of stay (d) Ventilator days a β is the coefficient of the variable from the logistic regression. Hypothermia = temperature 36.0 C; hypoxia = systemic oxygen saturation 88%. Long-term 6 months outcome is dichotomized into favorable good and unfavorable poor categories. predictive factors in our study is that we prospectively enrolled only those patients who were expected to survive the hospitalization. This study, therefore, was not designed to examine predictors of mortality. Patients with moribund head injuries who presented with fixed and dilated pupils or massive traumatic shock were excluded from the cohort. In addition, the pupillary examination was only abstracted from the chart in the ED assessments, and thus subjective variability to light reaction may have led to misclassifications. The initial GCS (13, 14, 18), LOC (13), requiring neurosurgery (14), LOS, and time on mechanical ventilation (8, 13, 20) have all been shown to be predictive of unfavorable outcome. We also found that children in the cmild/moderate head injury (GCS, 9 15) group had a high percentage of favorable long-term outcomes. In the cmild/moderate group, there were no deaths, and 98% of the patients sustained a favorable neurologic outcome, whereas in the severe TBI group, there were eight deaths, and 77% of patients had a favorable recovery. The association of an ICP monitor placement with a worse outcome should be viewed as a clear association, not a causative factor. For instance, ICP monitor placement is associated with worse outcomes, but this is likely reflective of a more serious brain injury and not the monitor placement itself, and therefore the association is not necessarily relevant to clinical decision making (13). ICP monitor placement was associated with an unfavorable outcome in our analysis, but was only significant in the bivariate regression (p < 0.01). ICP monitor placement was at the discretion of the attending neurosurgeon and severe patients who did not have an ICP monitor placed frequently improved in the ICU over the first 24 hours. The effect of seizures on outcome is unclear in prior studies, and it has been reported to negatively influence outcome (7, 21) or have no effect. In our analysis, the presence of seizures also emerged as predictive of unfavorable long-term neurologic outcome, but was only significant in the bivariate regression (p = 0.03). There is a paucity of pediatric data comparing status at discharge with long-term outcomes. Shaklai et al (13) in 2014 followed 77 patients for 10 years and showed that 61% functioned equally with their age-normative peers. However, of the 15 with unfavorable outcomes at discharge, only four improved. It has therefore been difficult to evaluate prognoses for children with unfavorable GOS or vegetative states at discharge. Our data from this large cohort, using the GOS and GOSEP, suggest that status at discharge does have a relationship to long-term outcome, but can change, as is shown by the considerable percentage of our patients that improved over time. As expected, patients who do well at discharge will likely continue to do so, with a high probability for developing a favorable long-term outcome. Importantly, this study also provides further support that many of the patients with unfavorable outcomes at discharge will go on to recover significant function. Remarkably, almost half of 6 XXX 2018 Volume XX Number XXX

7 Neurocritical Care the children who were discharged from the hospital with an unfavorable neurologic outcome recovered to improve outcome categories into a favorable category. Our results likely underrepresent the percentage of children who will have meaningful improvements in neurologic function after hospital discharge as even a one point score increase in the GOS or GOSEP may have a relevant impact on patients and their families. Although it is a relatively small sample size (n = 57), our results are among the first to describe that TBI at a younger age is not only associated with an unfavorable long-term outcome but that younger age was also associated in bivariate and multiple logistic regression analysis with remaining in an unfavorable outcome category if discharged with an unfavorable status. The eight patients who were vegetative at discharge are also a notable subgroup. All eight survived, seven regained consciousness and, strikingly, two emerged with a favorable outcome. One patient had a GOS of 4 after 8 months due to repeating a grade and requiring assistance with dressing. The other patient had a GOS of 5 after 6 months and per parental report was walking, running, and playing, with a tracheostomy removed and eating a regular diet, but this patient was still in rehabilitation and struggling with speech. The recovery to a favorable outcome in two of eight (25%) of this subpopulation is important to consider in the prognosis of patients who exhibit vegetative responses at discharge. It has been suggested that patients in a persistent vegetative state from trauma are unlikely to recover after 12 months (22). Although further research is needed in examining recovery potential after this time, these data suggest that it is possible for children who are classified as vegetative at hospital discharge to regain consciousness, and while less likely, could even progress toward a favorable long-term neurologic outcome. This study has a number of limitations. Care in the ED and in the ICU varied between patients, despite the publication of treatment guidelines in Introduction of these guidelines during the span of this study s data collection could have potentially impacted patient outcomes after their implementation in In this patient cohort, the primary aim was to examine long-term neurologic function rather than mortality. As such, the mortality rate is likely biased toward a lower rate due to the exclusion of children who were not expected to not survive. There is also the possibility of selection bias since only patients for whom written informed consent was obtained were included. When assessing predictive factors, variability in outcomes reflects the differences between specific inpatient therapies, the intensity of rehabilitation programs, and overall access to care. Although we dichotomized the long-term outcomes, two outcome measures were used over the study period to assess neurologic function, which could have affected the outcome categorizations. Neither the GOS, which was not widely available prior to 2005, nor the GOSEP capture adaptive deficits, increased risk of suicide, sleep-wake disturbances (23, 24), and Axis I and/or Axis II psychiatric disorders that are present in up to 50% of TBI survivors (25). The phrase, meaningful neurologic function in itself is a difficult phrase to define because it is truly specific to each patient and family. In this article, we defined outcome using dichotomous favorable and unfavorable outcome categories using the GOS, but important and potential relevant outcome characteristics may not be adequately represented using this categorization. Gross functional outcome categorization, while standard for reporting neurologic outcome, is inherently an over simplification of a complex self and family-based perception of recovery and quality of life. Prognostic data may help to identify injury or hospital variables that could aid clinicians and families in predicting favorable or unfavorable outcomes; however, they may fail to clearly define a more nuanced reality of what a family or individual considers a meaningful recovery. The decision making we refer to is extraordinarily difficult, and our analysis does not attempt to support decision making. However, consideration should be given to the observed improvement in outcome after hospital discharge, and to the proportion of children in the favorable outcome group within months of discharge. Our data illustrate how challenging long-term outcome prediction is in pediatric TBI, highlighting the need for future research. CONCLUSIONS This is the largest reported cohort to date of pediatric patients with cmild/moderate-to-severe accidental TBI with long-term follow-up and comparison to status at discharge. The age of the child at injury, ED GCS, mechanism of injury, and secondary traumatic events remain important factors in the determination of long-term neurologic function after accidental severe TBI pediatric patients. When status at discharge was compared with long-term outcomes, a substantial percentage of patients initially classified with unfavorable outcomes went on to have favorable recoveries at long-term follow-up. In addition, a majority of vegetative patients were observed to at least regain consciousness. Finally, further studies are needed to examine the notion that the developing brain may be particularly susceptible to diffuse traumatic injuries by impairing endogenous intrinsic repair and remodeling pathways. ACKNOWLEDGMENTS We thank a number of people involved in this study over the years. Initial funding came from the Crystal Charity Ball program for pediatric traumatic brain injury in conjunction with Dr. Sandi Chapman from The Center for Brain Health at The University of Texas at Dallas. We also thank Dr. Sam Lehman and Deborah Town of Children s Medical Center Dallas for the initial data collection and Katrina Vandebruinhorst for subsequent data collection and study coordination. In addition, Dr. Todd Maxson started the level 1 trauma center at Children s Medical Center Dallas, and Karen Dowling was the founding administrative coordinator for the Perot Center for Brain and Nerve Injuries. Ana Hernandez acquired the Glasgow Outcome Scale Extended Pediatrics as part of a larger outcomes project. Additional research coordinators participating in this study over the years include Patricia Plumb, Annemarie Jones, Evin Shirley, and Andrew Herbert. Assistance with statistical analysis was provided by Dr. Sekhar Ramkrishnan. Pediatric Critical Care Medicine 7

8 Slovis et al REFERENCES 1. Lingsma HF, Yue JK, Maas AI, et al; TRACK-TBI Investigators: Outcome prediction after mild and complicated mild traumatic brain injury: External validation of existing models and identification of new predictors using the TRACK-TBI pilot study. J Neurotrauma 2015; 32: Gagner C, Landry-Roy C, Lainé F, et al: Sleep-wake disturbances and fatigue after pediatric traumatic brain injury: A systematic review of the literature. J Neurotrauma 2015; 32: Richard YF, Swaine BR, Sylvestre MP, et al: The association between traumatic brain injury and suicide: Are kids at risk? Am J Epidemiol 2015; 182: Finnanger TG, Olsen A, Skandsen T, et al: Life after adolescent and adult moderate and severe traumatic brain injury: Self-reported executive, emotional, and behavioural function 2-5 years after injury. Behav Neurol 2015; 2015: Tompkins CA, Holland AL, Ratcliff G, et al: Predicting cognitive recovery from closed head-injury in children and adolescents. Brain Cogn 1990; 13: Adelson PD, Srinivas R, Chang Y, et al: Cerebrovascular response in children following severe traumatic brain injury. Childs Nerv Syst 2011; 27: Fulkerson DH, White IK, Rees JM, et al: Analysis of long-term (median 10.5 years) outcomes in children presenting with traumatic brain injury and an initial Glasgow Coma Scale score of 3 or 4. J Neurosurg Pediatr 2015; 16: Pfenninger J, Santi A: Severe traumatic brain injury in children are the results improving? Swiss Med Wkly 2002; 132: Tilford JM, Simpson PM, Yeh TS, et al: Variation in therapy and outcome for pediatric head trauma patients. Crit Care Med 2001; 29: Beers SR, Wisniewski SR, Garcia-Filion P, et al: Validity of a pediatric version of the Glasgow Outcome Scale-Extended. J Neurotrauma 2012; 29: Williams DH, Levin HS, Eisenberg HM: Mild head injury classification. Neurosurgery 1990; 27: Society of Critical Care M: Guidelines for the acute medical management of severe traumatic brain injury in infants, children, and adolescents. Crit Care Med 2003; 31:S407 S Shaklai S, Peretz R, Spasser R, et al: Long-term functional outcome after moderate-to-severe paediatric traumatic brain injury. Brain Inj 2014; 28: Chamoun RB, Robertson CS, Gopinath SP: Outcome in patients with blunt head trauma and a Glasgow Coma Scale score of 3 at presentation. J Neurosurg 2009; 111: Haider AH, Crompton JG, Oyetunji T, et al: Mechanism of injury predicts case fatality and functional outcomes in pediatric trauma patients: The case for its use in trauma outcomes studies. J Pediatr Surg 2011; 46: Walker WC, Ketchum JS 3rd, Marwitz JH, et al: Global outcome and late seizures after penetrating versus closed traumatic brain injury: A NIDRR TBI Model Systems Study. J Head Trauma Rehabil 2015; 30: Guerrier G, Morisse E, Barguil Y, et al: Severe traumatic brain injuries from motor vehicle-related events in New Caledonia: Epidemiology, outcome and public health consequences. Aust N Z J Public Health 2015; 39: Signorini DF, Andrews PJ, Jones PA, et al: Predicting survival using simple clinical variables: A case study in traumatic brain injury. J Neurol Neurosurg Psychiatry 1999; 66: Ramaiah VK, Sharma D, Ma L, et al: Admission oxygenation and ventilation parameters associated with discharge survival in severe pediatric traumatic brain injury. Childs Nerv Syst 2013; 29: DeCuypere M, Muhlbauer MS, Boop FA, et al: Pediatric intracranial gunshot wounds: The Memphis experience. J Neurosurg Pediatr 2016; 17: Eiben CF, Anderson TP, Lockman L, et al: Functional outcome of closed head injury in children and young adults. Arch Phys Med Rehabil 1984; 65: Multi-Society Task Force on PVS: Medical aspects of the persistent vegetative state (2). N Engl J Med 1994; 330: Anderson V, Catroppa C, Morse S, et al: Intellectual outcome from preschool traumatic brain injury: A 5-year prospective, longitudinal study. Pediatrics 2009; 124:e1064 e Huguenard AL, Miller BA, Sarda S, et al: Mild traumatic brain injury in children is associated with a low risk for posttraumatic seizures. J Neurosurg Pediatr 2015; 17: Velikonja D, Warriner E, Brum C: Profiles of emotional and behavioral sequelae following acquired brain injury: Cluster analysis of the Personality Assessment Inventory. J Clin Exp Neuropsychol 2010; 32: XXX 2018 Volume XX Number XXX