Sex Differences in the Outcomes of Mild Traumatic Brain Injury in Children Presenting to the Emergency Department

Abstract

Sex differences after concussion have been studied largely in high school and college athletes, often without reference to comparison groups without concussion. This study sought to evaluate sex differences in outcomes among all children and adolescents presenting to the Emergency Department (ED) for either mild traumatic brain injury (TBI) or orthopedic injury (OI), regardless of mechanism of injury. The study involved a concurrent cohort, prospective study design with longitudinal follow-up. Participants were eight to 16 years old with mild TBI (n = 143) or OI (n = 73). They were recruited and completed an initial assessment at EDs at two children's hospitals. They returned for a post-acute assessment within two weeks of injury and for follow-up assessments at three and six months. Outcomes included child and parent proxy ratings of somatic and cognitive symptoms, and standardized tests of cognitive functioning and balance. Sex did not moderate group differences in balance, fluid or crystallized cognitive ability, or child or parent proxy ratings of somatic or cognitive symptoms. Both parents and children reported more somatic symptoms in girls than boys, but in both groups. Compared with the OI group, the mild TBI group showed significantly lower fluid cognitive ability at the post-acute assessment and significantly higher somatic and cognitive symptoms according to both child and parent proxy ratings across the first two weeks post-injury. The results suggest that sex does not moderate the outcomes of mild TBI in a pediatric ED population. Previous research pointing to sex differences after concussion may reflect the lack of comparison groups, as well as a focus on adolescents and young adults and sport-related concussion. Future research should investigate whether sex moderates the outcomes of pediatric mild TBI in adolescents but not in pre-adolescent children.

Introduction

A concussion is a type of mild traumatic brain injury (TBI) that involves a disruption of normal brain functioning caused by a mechanical blow directly to the head or to the body with an impulsive force transmitted to the head.¹ Concussions in children and adolescents are particularly worrisome because of the vulnerability of the developing brain.^2,3 In the United States in 2013, among children younger than 15 years old, TBI accounted for more than 640,000 Emergency Department (ED) visits, most of which involved concussion or mild TBI.⁴ These numbers, however, are an underestimation of the true burden because they do not account for mild TBIs that are untreated or treated in non-ED settings.⁵

Most concussions in children and adolescents occur in sport and recreational settings.⁶ Estimates are that 1.1–1.9 million sport-related concussions occur each year among youth in the United States.⁷ Historically, research on sport-related concussion has focused on males given their participation in contact sports such as football. A growing number of females, however, have begun participating in youth sports.⁸ Although males sustain more concussions than females because of their greater exposure to higher-risk sports, females experience higher absolute injury rates for concussion after equating for exposure and sport.^9–11

Despite increasing female participation in sport and the heightened risk of concussion among female athletes, surprisingly little is known about possible sex differences in the outcomes of concussion.^12,13 Sex differences have been reported in a variety of outcomes, including post-concussive symptoms,^14,15 gait and balance,¹⁶ and cognitive test performance,^17–19 with females showing worse outcomes (e.g., more symptoms and slower symptom recovery) than males in most but not all studies.²⁰ In addition, female adolescent athletes may receive more interventions, including medication, vestibular therapy, and academic accommodations, than their male counterparts.²¹ Systematic reviews and meta-analyses also have confirmed worse outcomes in females after concussion,^22,23 although some reviews have suggested that sex differences do not remain significant when confounding variables such as baseline status are controlled.²⁴

Sex differences in the outcomes of concussion have been attributed to a variety of possible mechanisms. For instance, sex differences in symptom reporting might be attributable to differences in how females and males are socialized to admit or deny symptoms.²⁵ Physiological, metabolic, and neuroanatomical differences between males and females also might help account for sex differences after concussion. Sex-based differences in the brain's response to head trauma have been documented in neuroanatomy, cerebrovascular organization, and cellular response.^26–29 Males and females also have different levels of sex hormones, leading to suggestions that sex hormones may modulate the brain's response to concussion.^30,31 Finally, sex differences in concussion outcomes may arise from differences in neck strength and injury biomechanics.^32–35

Of course, these mechanisms are not all specific to concussion; for instance, sex differences in socialization could affect symptom reporting after any injury, or even in the absence of injury. Some of the purported mechanisms, however, are likely to be involved only in cases of concussion (i.e., brain's response to head trauma), such that girls would be differentially vulnerable after concussion versus injuries not involving the head.

Despite evidence of sex differences in concussion outcomes and plausible mechanisms for such differences, the existing research on pediatric concussion has a variety of significant gaps. Most of the research on sex differences has focused on sport-related concussion, typically among high school and college athletes. Very few studies have examined sex differences in concussions occurring outside sport or in younger school-age children.^15,16 In addition, many of the existing studies lack a comparison group of healthy individuals or individuals with injuries not involving the head, and instead simply compare males and females with concussion.^14,17,18 The absence of a comparison group, however, precludes any determination of whether sex actually moderates the effects of concussion versus simply accounting for variation in outcomes in a nonspecific fashion, irrespective of concussion.

The current study aimed to fill these gaps by examining sex differences in outcomes across the first six months post-injury among children ages 8–16 years presenting to the ED for mild TBI or orthopedic injury (OI). Outcomes included child and parent-proxy ratings of post-concussive symptoms and standardized tests of children's balance and cognitive functioning. We hypothesized that sex would moderate the outcomes of mild TBI, such that girls would show worse symptoms, balance, and cognitive functioning than boys after mild TBI, relative to any sex differences observed in the OI group.

Methods

Participants

The study involved a concurrent cohort, prospective, and longitudinal design that included children eight to 16 years old with mild TBI and a comparison group of children with mild OI not involving the head. Participants were recruited during initial visits to the EDs at Nationwide Children's Hospital in Columbus, Ohio, or Rainbow Babies and Children's Hospital in Cleveland, Ohio. Recruitment occurred from approximately 8:00 am to 10:00 pm, seven days a week, and lasted 46 months, from March 2014 to December 2017. All recruitment decisions regarding inclusion and exclusion criteria were reviewed by an attending physician at each site. The Institutional Review Boards at each site approved the study, and parents provided written informed consent and children provided written assent before any involvement in the study.

The mild TBI group included children who sustained a blunt head trauma that resulted in at least one of the following three criteria: (1) an observed loss of consciousness (LOC), (2) a Glasgow Coma Scale (GCS) score of 13 or 14,³⁶ or (3) at least two acute signs or symptoms of concussion as noted by ED medical personnel on a standard case report form (i.e., post-traumatic amnesia, focal neurological deficits, skull fracture, post-traumatic seizure, vomiting, headache, dizziness, other mental status changes). Exclusion criteria included delayed neurological deterioration, neurosurgical intervention, LOC >30 min, and GCS score <13. Children were not excluded if they were hospitalized or had intracranial lesions or skull fractures detected on CT scan.

Children were not required to complete a CT scan to be eligible for the study, but if CT scans were completed, they were available for the purpose of injury classification. Of the 40 children with mild TBI who had a CT scan, only four had trauma-related abnormalities. In addition, only six children demonstrated trauma-related abnormalities on magnetic resonance imaging (MRI) completed at the post-acute visit. Thus, the large majority of children in the mild TBI group had what would be diagnosed as a concussion.

The OI group included children who sustained upper or lower extremity fractures associated with Abbreviated Injury Scale (AIS) scores of 4 or less.³⁷ Exclusion criteria for the OI group included any head trauma or symptoms of concussion, as well as any injury necessitating surgery or the administration of sedative medication.

Children were not excluded from either group if they were administered analgesic medication in the ED, including narcotics. Additional exclusion criteria for both groups included: (a) any other severe injury as defined by an AIS score greater than 4; (b) any associated injury likely to interfere with neuropsychological testing (e.g., fracture of preferred upper extremity); (c) hypoxia, hypotension, or shock during or after the injury; (d) alcohol or drug ingestion involved with the injury; (e) history of previous TBI requiring hospitalization; (f) pre-morbid neurological disorder or mental retardation; (g) injury resulting from child abuse or assault; (h) history of psychiatric disorder requiring hospitalization; or (i) medical contraindication to MRI.

Of 588 eligible children (307 mild TBI, 281 OI) who were approached to participate in the study, 315 (195 mild TBI, 120 OI) consented and completed an initial assessment in the ED. Among those approached in the ED, those who consented did not differ in sex from those who declined, χ²(1) = 0.182, p = 0.669, or age, t(581) = 0.119, p = 0.905. Consent rates were higher in the mild TBI group (63.5%) than the OI group (42.7%), χ²(1) = 25.552, p < 0.001. In addition, consent rates varied by race, χ²(3) = 21.261, p < 0.001, being higher among Black (60.9%) and multi-racial (76.3%) children and lower in white (45.3%) children.

Those who consented also lived in census tracts with significantly lower median family income and percentage of residents below the poverty line than those who declined, p = 0.019 and p = 0.030 (Mann-Whitney U tests), respectively. Consent was not related to the percentage of minorities in children's census tracts, p = 0.216.

Of the 315 children who consented to participate, 217 (143 mild TBI, 74 OI) returned for a post-acute visit. Among those who consented, those seen only in the ED did not differ from those who returned for the post-acute visit on sex, χ²(1) = 0.078, p = 0.781, or age, t(313) = -0.257, p = 0.798. Participants in the mild TBI group (73.3%) were more likely to return for the post-acute visit than participants in the OI group (61.7%), χ²(1) = 4.718 p = 0.030.

Return rates for the post-acute visit also varied by race, χ²(3) = 32.579, p < 0.001, with multi-racial (96.6%) and white (79.8%) children more likely to return than Black children (54.3%). Children who returned also resided in census tracts with higher median family incomes and lower percentages of minorities and residents below the poverty line than those who were seen only in the ED, p = 0.003, < 0.001, and = 0.028 (Mann-Whitney U tests), respectively.

The core sample for this study included the 143 children with mild TBI and 74 children with OI who returned for the post-acute visit. Participants who returned for the post-acute visit did not differ significantly from the combined group of children who declined to participate in the study or were seen only in the ED in terms of age, t(581) = -0.011, p = 0.991, sex, χ²(1) = 0.248, p = 0.618, or census tract measures of median family income and percentage of minorities and residents below the poverty line, all p > 0.451 (Mann-Whitney U tests). They did differ in race, χ²(3) = 23.594, p < 0.001, largely because a higher percentage of multi-racial children returned for the post-acute visit (73.7%) than did white (36.1%) or Black (33.1%) children.

As summarized in Table 1, the mild TBI and OI groups did not differ based on sex, age at injury, or Full Scale IQ. The mild TBI and OI groups did differ in race, χ²(3) = 8.297, p = 0.040; the mild TBI group had a higher percentage of white participants (52.4%) and lower percentage of multi-racial (9.8%) participants than the OI group (37.8% and 18.9%, respectively). In addition, the mild TBI group had higher socioeconomic status (SES) than the OI group, t(215) = 2.794, p = 0.006, as measured by a z-score composite of standardized measures of total years of maternal education, median family income for the family census tract, and occupational status.³⁸ Of the 217 children completing the post-acute visit, 159 (102 mild TBI, 57 OI) returned at three months and 143 (97 mild TBI, 46 OI) returned at six months. The mild TBI and OI groups did not differ in attrition rates at three or six months.

Table 1.

Demographic and Clinical Characteristics of Participants

Variable	OI (n = 74)	Mild TBI (n = 143)
Age at injury, y, mean (SD)	12.35 (2.37)	12.56 (2.61)
Male, No. (%)	48 (64.9)	95 (66.4)
Race, No. (%),
White	28 (37.8)	75 (52.4)
Black	32 (43.2)	50 (35.0)
Multi-racial	14 (18.9)	14 (9.8)
Other	0 (0.0)	4 (2.8)
SES composite z score, mean (SD)	-0.26 (0.90)	0.13 (1.03)
Full-scale IQ, mean (SD)	97.89 (15.54)	98.71 (15.04)
Loss of consciousness, No. (%)1	-	38 (27.1)
Glasgow Coma Scale score <15, No. (%)2	-	15 (11.7)
Injured in sport or recreational setting, No. (%)3	55 (80.9)	110 (78.0)
Injury mechanism
Fall	32 (43.8)	47 (33.3)
Struck object	18 (24.7)	37 (26.2)
Struck person	7 (9.6)	36 (25.5)
Bicycle or motor vehicle related	7 (9.6)	21 (14.9)
Other	9 (12.3)	0 (0.0)

OI, orthopedic injury; TBI, traumatic brain injury; SD, standard deviation; SES, socioeconomic status; IQ, intelligence quotient.

¹Data missing n = 3.

²Data missing n = 15.

³Data missing n = 9.

Procedure

Children completed an initial assessment at the time of recruitment, during their initial visit to the ED, that included a balance measure, as well as retrospective parent ratings of children's pre-injury symptoms. Children did not provide retrospective ratings of pre-injury symptoms. Participants and their families returned for post-acute visits, within three weeks of injury (mean = 10.31 days, standard deviation = 2.84, range 3–18 days), and also for follow-up assessments at three and six months post-injury.

All post-injury assessments included self-ratings of post-concussive symptoms by children and proxy ratings by parents, as well as balance and cognitive testing of the children. Children also completed a performance validity measure at all post-injury assessments, as well as a measure of intellectual functioning at the post-acute assessment. In addition, parents and children were asked to complete weekly symptom ratings remotely using a phone app or REDCap,³⁹ beginning the week after the post-acute assessment.

Measures

Pre-morbid intellectual functioning

The two-subtest version of the Wechsler Abbreviated Scale of Intelligence (WASI) was used to assess the children's general intellectual functioning and provided an estimated Full Scale IQ.⁴⁰

Acute signs/symptoms

A standardized injury case report form was used to elicit descriptive information regarding the injury and acute signs and symptoms of concussion including LOC, GCS scores, mechanism of injury, neurological status, reports of retrograde or anterograde amnesia, and other clinical factors. The information was gathered from medical records and medical personnel by recruiters and verified by attending physicians.

Balance

The Balance Error Scoring System (BESS) was used to assess three stances (narrow double leg stance, single leg stance, and tandem stance) and two footing surfaces (firm surface/floor vs. medium density foam) held for 20 sec with hands on hips and eyes closed. Error points, up to 10 per trial, were given for opening eyes, lifting hands off hips, and stepping, stumbling, or falling. The BESS has shown satisfactory reliability in children and adolescents.⁴¹

In the current study, the administration of the BESS was videotaped, and approximately 10% of the assessments were scored twice by independent raters. The BESS total score showed excellent reliability as assessed using the Krippendorf alpha (0.953), with satisfactory reliability on each of the six trials (alpha = 0.64–0.86). Because of the nature of children's injuries (e.g., lower extremity fracture) or technical problems (e.g., incomplete video coverage of the body), 30.8% of the mild TBI group and 63.5% of the OI group did not have BESS data from their ED visit; in addition, 48.6% of the OI group did not have BESS data from the post-acute visit.

More than 90% of children with mild TBI or OI had BESS data at all other assessments; the number of children with available BESS data at each assessment ranged from 89 to 129 for the mild TBI group and from 27 to 52 for the OI group. Because analyses used all available data, however, only 10 children from the core sample, two with mild TBI and eight with OI, were excluded entirely from analyses of the BESS.

Cognitive functioning

The National Institutes of Health (NIH) Toolbox Cognition Battery (CB) was used to assess cognitive abilities.⁴² The battery consists of seven individual subtests that assess cognitive flexibility, visual attention and inhibitory control, processing speed, episodic memory, working memory, vocabulary knowledge, and reading. Standard scores corrected for age were derived for the fluid and crystallized cognitive composites, which were used as outcomes for this study.

Post-concussive symptoms

Post-concussive symptoms were assessed using the Health and Behavior Inventory (HBI).⁴³ The HBI consists of 20 items that describe a variety of cognitive and somatic symptoms, yielding two subscales derived from factor analysis. Participants rate the frequency of occurrence of each item on a 4-point scale (0 = never, 1 = rarely, 2 = sometimes, 3 = often). The HBI includes child and parent proxy forms, worded slightly differently to reflect first versus third-person perspectives. The HBI is a core measure in the Common Data Elements for Pediatric Traumatic Brain Injury and incorporated in the Child Sport Concussion Assessment Tool, 5th edition.^44,45

The HBI has shown satisfactory reliability and validity in previous research.^43,46 In the current sample, the Cronbach alpha was above 0.80 for the somatic scale and above 0.90 for the cognitive scale for both child and parent proxy ratings at each of the scheduled visits.

Medical Symptom Validity Test (MSVT)

Children's performance validity was measured using the MSVT, which is a computerized, forced-choice recognition test designed to detect invalid performance. Children eight years and older are generally able to perform at very high levels of recognition accuracy, commensurate with adults.⁴⁷ Performance on the MSVT accounts for substantial variance in cognitive test performance among children with mild TBI.^48,49

The traditional MSVT cutoff was used initially to classify responses as being potentially invalid (i.e., a score of 85% or less on all three easy subtests). Of the participants meeting that criteria, some displayed a pattern of substantially worse performance on the difficult subtests than on the easy ones (difference of >20%), suggestive of cognitive impairment or memory deficit.⁴⁷ Only participants who met the traditional MSVT cutoff but did not display evidence of cognitive impairment or memory deficit were considered to have potentially invalid performance.

Data analysis

Linear mixed models were conducted for the measures of balance (i.e., BESS total score) and cognitive abilities (i.e., NIH Toolbox CB fluid and crystallized cognition composite age-adjusted standard scores), which had up to four (balance) or three (cognitive functioning) assessments available for analysis. Mixed models are advantageous with longitudinal data because they use all available outcome data for estimation. For each outcome, a fixed effects linear mixed model was fit with sex, group, days since injury, and their interaction as predictors. The SES also was included as a covariate in all analyses, and the model for the BESS total score also included age at injury as a covariate; the two cognitive abilities measures were already adjusted for age.

Preliminary analyses indicated that symptom ratings for most children with mild TBI had returned to pre-injury levels by four weeks after injury, and also that weekly ratings became increasingly sparse after around 14 weeks post-injury. Therefore, the dataset that we modeled included ratings from the post-acute and three-month face-to-face assessments, as well as all weekly ratings out to 13 weeks after the post-acute assessment.

We fit generalized least squares models with a continuous first-order autoregressive structure to allow for correlated residuals from repeated measurement over time (i.e. nonindependent data).⁵⁰ For each symptom scale, a model was fit with group, sex, days since injury, and their interactions as predictors. Age at injury, SES, and parents' retrospective ratings of pre-injury symptoms were included as covariates. To allow for nonlinearity, flexible restricted cubic splines were applied to all continuous predictors (age, pre-injury symptom rating, SES z-score, and time).⁵¹

For all predictors except for time, four knots (i.e., pivot points) were afforded at strategically spaced quantiles (i.e., 0.05, 0.35, 0.65, 0.95). For time, we specifically afforded five front-loaded knots (i.e., at 2, 3, 4, 8, 12 weeks) to accommodate our expectation for nonlinearity, particularly during the early phase of recovery. The generalized least square models also used all available outcome data for estimation. Significant group differences were explored using post-model fit contrasts by computing estimates of the Cohen d by dividing differences between estimated means by context-specific standard deviations calculated from the study sample.

For both linear mixed model and generalized least square analyses, if an interaction did not approach statistical significance (p > 0.10) after initial fitting, the model was refitted without the interaction term (starting with the highest-order term) until only significant interaction terms remained in the final model. Final models were also reestimated after excluding 10 children (nine mild TBI, one OI) who demonstrated invalid MSVT performance at any post-injury assessment. Children with suspected cognitive impairment or memory problems (seven mild TBI, two OI) were retained in these analyses.

Results

Balance

The three-way interaction of group, sex, and time since injury was significant for the BESS total score, F(1,156) = 5.35, p = 0.021. Post-model fit contrasts, however, showed no significant differences at any visit between the mild TBI and OI groups within either sex, or between boys and girls within either group. Tests of trends over time showed that the girls with mild TBI showed a significant decline in errors over time, t(114) = -3.44, p < 0.001, but no other groups showed a significant change in errors over time. In the full model, higher SES and older age both predicted significantly lower total scores, F(1,198) = 10.69, p = 0.001, and F(1,204) = 30.11, p < 0.001, respectively.

The results did not change when children who demonstrated any invalid MSVT performance were excluded and the model reestimated. Notably, though, after removing one girl with OI who displayed an atypical pattern of increasing errors over time, the three-way interaction was no longer significant, nor were any effects involving group or sex.

Cognitive functioning

For the NIH Toolbox CB fluid cognition composite, the three-way interaction of group, sex, and time since injury did not approach significance, F(1,172) = 0.25, p = 0.615, and was trimmed. The two-way interactions of sex and time and sex and group also did not approach significance and were trimmed. The interaction of group and time approached significance, F(1,167) = 3.75, p = 0.055. Follow-up tests showed that the mild TBI group showed significantly lower fluid cognition composite scores at the post-acute visit, but not at three or six months (see Fig. 1).

FIG. 1.

Model estimates (M, 95% confidence interval) of performance on the National Institutes of Health Toolbox Cognition Battery fluid cognition composite by group. TBI, traumatic brain injury; OI, orthopedic injury.

Tests of trends showed that both groups showed significant increases in fluid cognition scores over time, t(164) = 7.22, p < 0.001 and t(168) = 2.72, p = 0.007, for mild TBI and OI groups, respectively—these increases likely reflecting practice effects. In the final model, higher SES predicted significantly higher fluid cognition scores, F(1,210) = 14.75, p < 0.001. The results did not change when children who demonstrated any invalid MSVT performance were excluded and the final model reestimated.

For the crystallized cognition composite, the three-way interaction of group, sex, and time since injury did not approach significance, F(1,175) = 1.44, p = 0.232, and was trimmed. The interactions of group and time and sex and time also did not approach significance and were trimmed. The interaction of group and sex was significant, F(1,225) = 3.97, p = 0.048. Follow-up tests showed girls had significantly lower scores than boys in the OI group. No significant differences were found, however, between the mild TBI and OI groups within either sex or between boys and girls in the mild TBI group. The main effect of time was not significant, F(1,166) = 2.49, p = 0.116, suggesting that neither group showed a significant change in the crystallized cognition composite over time.

In the final model, higher SES predicted significantly higher crystallized cognition composite scores, F(1,215) = 68.48, p < 0.001. After children with any invalid MSVT performance were excluded and the final model reestimated, the interaction of group and sex approached significance, F(1,214) = 3.60, p = 0.059, and follow-up tests followed the same pattern as those using the full sample.

Post-concussive symptoms

Child ratings

For children's ratings of somatic symptoms, the three-way interaction of group, sex, and time since injury did not approach significance, χ²(4) = 4.96, p = 0.291, and was trimmed. The two-way interactions of sex and time and sex and group also did not approach significance and were trimmed. The interaction of group and time since injury was significant, χ²(4) = 35.69, p < 0.001. Post-model fit contrasts showed that the mild TBI group reported significantly higher somatic symptom ratings at one and two weeks, but not at four, eight, or 12 weeks (see Fig. 2a). In the final model, girls reported higher somatic ratings than boys, albeit not significantly, χ²(1) = 3.75, p = 0.053.

FIG. 2.

Model estimates (M, 95% confidence interval) of ratings on the Health and Behavior Inventory Child (HBI-C) and Parent (HBI-P) forms. (a) Child self-ratings of somatic symptoms. (b) Parent proxy ratings of somatic symptoms. (c) Child self-ratings of cognitive symptoms. (d) Parent proxy ratings of cognitive symptoms. OI, orthopedic injury; TBI, traumatic brain injury.

Higher SES predicted significantly lower somatic symptoms, χ²(4) = 11.55, p = 0.021. Higher parent ratings of pre-morbid somatic symptoms also were associated with higher post-injury somatic symptom scores, but not significantly, χ²(3) = 7.08, p = 0.069. The results did not change when children who demonstrated invalid MSVT performance were excluded and the final model reestimated, except that girls now reported significantly more somatic symptoms than boys, χ²(1) = 6.86, p = 0.009.

For children's ratings of cognitive symptoms, the three-way interaction of group, sex, and time since injury did not approach significance, χ²(4) = 1.79, p = 0.774, and was trimmed. The two-way interactions of sex and time and sex and group also did not approach significance and were trimmed. The interaction of group and time since injury was significant, χ²(4) = 13.95, p = 0.007. Post-model fit contrasts showed that the mild TBI group reported significantly higher cognitive symptom ratings at one and two weeks, but not at four, eight, or 12 weeks (see Fig. 2c).

In the final model, higher SES predicted significantly lower somatic symptoms, χ²(4) = 10.61, p = 0.031, as did higher parent ratings of pre-morbid cognitive symptoms, χ²(4) = 40.77, p < 0.001. The results were unchanged when children who had any invalid MSVT performance were excluded and the final model reestimated.

Parent proxy ratings

For parent proxy ratings of children's somatic symptoms, the three-way interaction of group, sex, and time since injury did not approach significance, χ²(4) = 1.98, p = 0.740, and was trimmed. The interaction of group and time since injury was significant, χ²(4) = 40.58, p < 0.001, but no interactions involving sex approached significance, and the two-way interactions of sex and time and sex and group were trimmed. Post-model fit contrasts showed that parents reported significantly higher somatic symptom ratings in the mild TBI group at one and two weeks, but not at four, eight, or 12 weeks (see Fig. 2b).

In the final model, parents reported significantly higher somatic symptoms in girls than boys across groups, χ²(1) = 5.87, p = 0.015. Higher parent ratings of pre-morbid somatic symptoms also predicted significantly higher post-injury somatic symptoms, χ²(3) = 65.54, p < 0.001. When children who demonstrated invalid MSVT performance were excluded and the final model reestimated, the results were essentially unchanged.

For parent proxy ratings of children's cognitive symptoms, the three-way interaction of group, sex, and time since injury did not approach significance, χ²(4) = 2.36, p = 0.670, and was trimmed. None of the two-way interactions involving sex approached significance, so they were trimmed as well. The interaction of group and time since injury was significant, χ²(4) = 9.58, p = 0.048, as was the overall effect of group, χ²(5) = 11.23, p = 0.047. Post-model fit contrasts showed that parents reported significantly higher cognitive symptom ratings in the mild TBI group at one and two weeks, but not at four, eight, or 12 weeks (see Fig. 2d).

In the final model, higher parent ratings of pre-morbid cognitive symptoms predicted significantly higher post-injury cognitive symptoms, χ²(4) = 241.90, p < 0.001, as did older age, χ²(4) = 13.73, p = 0.008. When children who had any invalid MSVT performance were excluded and the final model reestimated, the main effect of group and the interaction of group and time were no longer statistically significant, χ²(5) = 9.36, p = 0.096, and, χ²(4) = 8.24, p = 0.083. Follow-up tests of group differences, however, followed a similar pattern as those in the complete sample.

Discussion

Given the growing recognition of the public health burden posed by pediatric concussion and the possibility of sex differences in outcomes, this study sought to evaluate sex as a possible moderator of outcomes among children and adolescents presenting to the ED for mild TBI. Although girls were hypothesized to experience poorer outcomes than boys after mild TBI relative to OI, the results did not support that hypothesis. Instead, the results indicate that sex did not moderate group differences in balance, fluid or crystallized cognitive ability, or child or parent proxy ratings of somatic or cognitive symptoms.

Both children and parents reported higher somatic symptoms for girls than boys, but that difference was true regardless of injury type. Thus, no differential effect of concussion by sex was apparent across a range of outcomes. In contrast, compared with the OI group, the mild TBI group showed significantly lower fluid cognitive ability at the post-acute assessment and significantly higher somatic and cognitive symptoms according to both child and parent proxy ratings across the first two weeks post-injury.

Our results appear to contradict previous research suggesting sex differences in outcomes specific to concussion. Several explanations may help account for the discrepancy in findings. First, our study included an OI comparison group, which allowed us to assess the effect of concussion relative to other injuries. Many other studies that have found sex differences after concussion did not include control groups, precluding attribution of sex differences to concussion specifically.^14,15,21

Second, our study focused on children and adolescents who were injured in a variety of settings, whereas most previous research on sex differences has focused on adolescents and young adults with sport-related concussion.^{14,17–19,24} We controlled for age in our analyses, but did not have a large enough sample to provide sufficient statistical power to test whether sex is a moderator of concussion outcomes in adolescents but not in pre-adolescent children, or whether sex differences are more apparent after concussions occurring in sport versus non-sport settings.

Third, our study population included a higher proportion of nonwhite children than previous studies. Black children have demonstrated less knowledge about concussion signs and symptoms than their non-Hispanic white counterparts.⁵² In addition, Black children are 30% less likely to receive a diagnosis of concussion in the ED than non-Black children when presenting to the ED with a head injury.⁵³ These findings may be an indication both that Black children report their symptoms differently than white children and that healthcare providers interpret and treat their symptoms differently. More research is needed to better understand how race and ethnicity may be related to how individuals perceive and report their concussion symptoms.

Finally, our study evaluated multiple outcomes, including both subjective symptom ratings and objective measures of balance and cognition. By contrast, most previous studies have focused on symptoms.^15,24 We found a sex difference in child and parent proxy ratings of somatic symptoms, but that was true of both injury types, again highlighting the importance of including an OI comparison group. Somatic symptoms are also reported more commonly by girls than boys in uninjured samples, particularly during adolescence, suggesting that the differences found in this study are not necessarily a consequence of the children's injuries.^54–56

Girls had significantly lower crystallized cognitive ability than boys in the OI group, but sex did not moderate group differences in that outcome. Girls with mild TBI showed a decrease in balance errors over time that was not apparent in girls with OI or boys with either OI or mild TBI; again, though, sex did not significantly moderate group differences.

Our findings regarding group differences in outcomes are much more consistent with previous research. Systematic review and meta-analyses, together with large individual studies, have shown that children with mild TBI, as a group, demonstrate deficits in certain cognitive abilities, especially processing speed and executive functions, as well as heightened post-concussive symptoms, that are most apparent post-acutely and tend to resolve within one month of injury.^57,58

We were somewhat surprised not to find evidence of acute or post-acute balance difficulties after mild TBI, given reports of such differences in studies of sport-related concussion.^59–61 Our comparison group of children with OI, however, may also have been vulnerable to balance problems given the nature of their injuries, especially on acute presentation to the ED. Indeed, many children in the OI group and some in the mild TBI group were unable to complete the BESS acutely or post-acutely because of their injuries. This also reduced the sample size available and potentially biased the comparisons, although only 10 children were excluded entirely from analyses of the BESS. In future articles, we will examine the nature and predictors of the effects of mild TBI more extensively using data from the larger parent project.

Our study has several other limitations. The study was conducted at two EDs in children aged eight to 16 years old. Therefore, the results cannot be generalized to children with mild TBI or OI who do not seek emergency care or to children outside this age range. In addition, recruitment occurred between the hours of 8 am and 10 pm, and children seen for mild TBI or OI during the off hours may have differed from those seen during the hours of recruitment. In a previous project, however, that was conducted at the same sites and used similar inclusion and exclusion criteria,⁴⁶ 89% of participants presented to the ED during the 8 am–10 pm interval and the participants who arrived at the ED during this interval did not differ from those who arrived during the 10 pm–8 am interval on demographics, injury types/severity, or post-injury outcomes.

The second limitation concerns statistical power. While samples sizes for the female mild TBI, male mild TBI, and male OI groups ranged from 48 to 95, the female OI group had only 26 participants, limiting statistical power to detect interactions with sex. In addition, attrition was a concern given the longitudinal study design. Analyses, however, suggested limited bias because of attrition; children who completed post-acute assessments were similar to the combined group of children who either did not consent in the ED or who consented but did not return for post-acute assessments. Moreover, attrition rates were similar in both groups after the post-acute visit, and analyses used all available outcome data.

Another limitation is that the observational study design precludes definite conclusions about causality. Our assumption, however, is that group differences in outcomes, as well as any differential effects by sex, are related to mild TBI because the groups were similar demographically and also on retrospective ratings of pre-injury symptoms and functioning. A final limitation is that retrospective ratings of children's pre-injury symptoms were only obtained from parents. Given the limited agreement between parent and child symptom ratings,^62,63 future research may benefit from obtaining retrospective ratings directly from children as well.

Despite these limitations, our findings with regard to sex differences may have clinical implications. Although the results show no sex differences in the effects of concussion, significant sex differences were found in both the mild TBI and OI groups on both child and parent ratings of somatic symptoms. In clinical practice, this is likely to translate into higher somatic symptom reports among girls than boys with mild TBI.

Healthcare providers who work with children with mild TBI need to be aware of potential gender biases in clinical practice and not assume that girls are overreporters who dramatize somatic symptoms or that boys are underreporters who suppress those same symptoms. Simply put, healthcare providers need to be aware that symptom reporting, whether by children or by their parents and independent of injury type, is influenced by many factors, including sex, and take those factors into account in their clinical management.

Conclusion

This represents one of the first studies to evaluate sex differences in the outcomes of mild TBI, compared with OI, among a pediatric ED population. The study examined multiple outcomes and found evidence for effects of mild TBI on fluid cognitive abilities and somatic and cognitive symptoms in the first two weeks post-injury, but did not find differential effects of mild TBI by sex in any domain. The findings suggest only limited sex differences in the outcomes of concussion in this population.

Future research is needed in larger samples to determine whether sex is a moderator of the outcomes of mild TBI specifically during adolescence, as opposed to earlier in childhood. Given the suggestion that sex hormones may modulate the effect of concussion,^30,31 future research also should explore the possible role of pubertal status as opposed to age as a moderator of outcomes in both boys and girls.

Funding Information

This work was supported by the National Institutes of Health grant R01HD076885.

Author Disclosure Statement

No competing financial interests exist.