Articles were identified through online database searching, hand-searching reference lists (i.e., pearl growing), and performing cited reference searches (see Figure 1). Databases included PubMed, MEDLINE^®, PsycINFO^®, CINAHL, Cochrane Library, EMBASE, SPORTDiscus, Scopus, and Web of Science. Articles published in English from database inception to May 15, 2021 were included in our searches (Registration: PROSPERO CRD42016041479, CRD42019128300).

Two general categories relating to (i) sport and athlete-related terms, and (ii) brain concussion related terms were used as the key search terms, as follows: sport, sports [MeSH], athletic, athlete, player AND craniocerebral trauma, brain injuries, brain concussion, sports concussion, athletic injuries, mild traumatic brain injury, mTBI, traumatic brain injury, TBI, brain concussion, concussion, multiple concussions, repeated concussion, repetitive concussion, cumulative concussions, concussion history, brain damage, prognosis, outcome, recovery, risk factor, injury incidence, sex differences, gender, genetics, ApoE, BDNF, S100B, GFAP, severity, loss of consciousness, LOC, post-traumatic amnesia, PTA, amnesia, retrograde amnesia, seizure, seizures, learning disorder, ADHD, level of education, migraine, mental health, sleep disorders, medications, cervical injury, vestibular injury, psychological reactions, anxiety, depression, headaches, intractable headaches, magnetic resonance imaging, MRI, computer tomography, and CT. This extraordinarily broad set of search terms was used to replicate the terms used in our prior systematic reviews (12–16) because many articles examine multiple predictors, and race/ethnicity might simply be included as a secondary or demographic variable, and thus, a more focused search would likely miss articles that did in fact examine race/ethnicity as one of many predictors or variables analyzed.

To be included, clinical outcomes needed to be measured one or more weeks following injury. We did not study acute clinical outcome. Outcome and clinical recovery were defined broadly to include self-reported resolution of concussion symptoms, improvement in neuropsychological functioning, improvement in vestibular deficits, and time to medical clearance and return to play and/or participation in normal activities, such as school.

Three authors (AKK, CG, and BL) completed the extraction of information from the articles [i.e., first author, year of publication, the PubMed identification number, number of participants, proportion of female participants, age, setting, time since injury, characteristics of the index concussion (e.g., proportion experiencing loss of consciousness, posttraumatic amnesia), study design, outcome examined, and the period for assessing the clinical outcome]. Authors extracted details regarding the association between race and outcome from the index concussion, including results of statistical significance testing, direction of effects, and any quantitative metrics reported in the study (e.g., recovery times, parameter estimates from statistical modeling, effect sizes). Discrepancies were resolved by discussion.

For the included studies, three authors (AKK, CG, and BL) coded whether social determinants of health (SDoH) and associated subcategories were reflected. A coding sheet was developed by the study team to identify SDoH and associated subcategories derived from both the Healthy People 2020 (https://www.healthypeople.gov/2020/topics-objectives/topic/social-determinants-of-health) and the Healthy People 2030 websites (https://health.gov/healthypeople/objectives-and-data/social-determinants-health). The five SDoH domains identified by Healthy People were included: Economic Stability, Education Assess and Quality, Health Care Access and Quality, Neighborhood and Built Environment, and Social and Community Context. Healthy People 2020 provided a list of several key issues and underlying factors associated with each of the five domains. For example, an important issue included under Economic Stability is “poverty.” All the key issues and underlying factors were included as separate variables to code, and the list was cross-referenced with material on the Healthy People 2030 website, where a separate webpage is available for each SDoH domain. Three co-authors (NC, AKK, and BL) read the text descriptions and identified other key issues or underlying factors that were not included in the Healthy People 2020 bulleted list.

The same three authors (AKK, CG, and BL) also collected information about the sample size, average age, age range, gender, racial composition, and ethnic composition of the sample. The extent to which the studies analyzed or provided information about key health equity variables (e.g., culture/acculturation, socioeconomic status, and language) was characterized as follows: the study (i) provided no mention of the variable, (ii) included the variable as a demographic category only, or (iii) examined the variable in depth (e.g., the health equity variable was a primary variable of interest in the study, or outcome or prognostic results were stratified and reported across levels of the variable). We also coded race and ethnicity in this manner. The content analysis coding sheet is included in the Supplementary material.

Risk of bias for each article was rated by three authors (NC, AKK, and CG) using the Newcastle-Ottawa Scale, a risk of bias assessment system designed for observational studies (17). Differences in risk of bias ratings were resolved by discussion. The scale includes nine items assessing specific study features relating to Sample Selection, Comparability between Research Groups, and Outcome (for Cohort studies) or Exposure (for Case Control studies). For each of the nine items, credit is assigned for stronger, higher quality design features based on criteria presented on the scoring sheets and in the coding manual (17). This scale has been reported to be suitable for systematic reviews (18). For the current review, when rating the Newcastle-Ottawa Scale, we considered race or ethnicity as the “exposure” and prolonged recovery or worse clinical outcome (e.g., symptom duration ≥28 days, worse neurocognitive test performance, longer duration of return to play) from the index concussion as the “outcome.” Additionally, the level of evidence for each article was rated by the same three authors (NC, AKK, and CG) using the Oxford Classification for Evidence-Based Medicine. Differences in the level of evidence ratings were resolved by discussion.

Substantial heterogeneity in the participant populations, comparators, clinical outcomes, and recovery time periods precluded quantitative synthesis and analyses. We performed a narrative synthesis and descriptively summarized the study samples and methodologies. We summarized the results of statistical significance testing (using each study authors' definition of significance, most often was p < 0.05) regarding whether clinical outcome from concussion was or was not associated with race or ethnicity, including the direction of effects. We reported the results of our content analysis descriptively.

A total of 5,118 abstracts were screened, 619 full-text articles were evaluated for inclusion, and 12 studies met inclusion criteria (19–30). Four studies (25%) reported concussion outcome among children or adolescents (maximum sample age ≤ 19 years) (19–22). Eight studies (75%) reported outcome among college-age samples (i.e., the mean sample age >18, they included participants age ≥ 20 years, or they indicated the sample consisted exclusively of college athletes) (23–30). Characteristics for the study populations are summarized in Table 1. There were 2,887 participants (mean age = 17.8 years, SD = 2.04, median = 18.9 years among the eight studies that reported sample age). Two studies involved overlapping samples (19, 20)—the number of non-overlapping participants was 2,769. Ten (83.3%) of the studies were published since 2015, and 5 (41.7%) were published since 2018. One study was published in 2010 and one study was published in 2004. Studies included a median of 26.0% girls (range 0%-50.8%). The racial and ethnic compositions of the included studies are summarized in Table 2. Two-thirds of the studies (66.7%) were prospective designs and one-third (33.3%) were retrospective. Studies in this review examined the following outcomes: recovery time/symptom duration (k = 6) (21, 22, 24–26, 30), time to return to sports (k = 5) (19, 20, 26, 27, 30), mental health problems (k = 2) (28, 29), and neuropsychological functioning and symptom severity (k = 1) (23) following concussion [several studies reported multiple outcomes, so the number of studies sums to greater than 12].

Level of Evidence and Newcastle-Ottawa Scale ratings are summarized in Table 3. The mean level of evidence was 3.3 (SD = 0.5). Nine studies (75%) were rated as level 3 (i.e., cohort studies) (19, 20, 22, 24–27, 29, 30) and 3 studies (25%) as level 4 (i.e., case-control studies or lower-quality prognostic cohort studies) (21, 23, 28). Studies received credit for between three to eight Newcastle-Ottawa Scale items (mean = 5.7, SD = 1.3).

Across all studies and outcomes, the most frequent finding was that clinical outcomes did not differ based on race or ethnicity. Most studies (58.3%) did not find a statistically significant association between race or ethnicity and outcome from concussion. Three studies (25.0%) reported a statistically significant result (20, 23, 28) and the remaining two studies (16.7%) reported predominantly null results with an isolated statistically significant finding (22, 26). Of note, among these studies with at least one statistically significant result, two (40%) reported that racial and ethnic minority youth had better outcomes (i.e., faster recovery times, less likely to take >4 weeks to recover) (20, 22). The remaining three studies reported that racial and ethnic minority youth had worse outcomes (i.e., worse subacute neurocognitive test performance, higher levels of depressive symptoms, and longer symptom resolution time) (23, 26, 28). Among the included articles, only three studies (25%) were specifically designed and carried out to examine whether race and ethnicity were associated with clinical outcomes following concussion (19, 20, 23), and two of these studies reported statistically significant results (20, 23) while the other study reported a null result. However, there are major caveats and considerable nuance to appreciate and consider when interpreting results from these studies. Further detail about the individual studies and their findings are provided in sections below.

Four studies were identified that examined the potential association between race and outcome from a sport-related concussion in children and adolescents (19–22), and two of those studies reported on a partially overlapping sample (19, 20). Thus, there were three independent samples totaling 2,080 youth studied. All three pediatric samples were recruited from specialty concussion clinics.

The clinical outcome in all four studies was recovery time. This included days to recovery for two studies (19, 20), recovery within several time periods for one study (i.e., within 2 weeks, 3 weeks, 4 weeks, or >4 weeks) (22), and persistent symptoms lasting >3 months following an injury in the fourth study (21). Two studies with partially overlapping samples reported mixed results. An initial study of 118 youth noted a 4 day difference in median recovery time between youth identified as White race (n = 72, median 17 days) compared to youth identified as “Minority” race (n = 46, median 13 days), and this difference was not statistically significant in either univariate (p = 0.06) or multivariate (p = 0.50) analyses (19). The investigators added 109 additional youth to the sample and replicated the analyses. With the larger sample (N = 227), the authors noted an eight day difference in median recovery time between youth identified as White race (n = 122, median 19 days) compared to youth identified as a combined Hispanic/African American group (n = 105, median 11 days), and this difference was statistically significant in univariate analyses (p < 0.001) (20). In multivariate models, the authors further stratified their racial and ethnic analyses by gender. Compared to girls who identified as White race (the reference group in these analyses), neither girls who identified as Hispanic or African American (combined; p = 0.91) nor boys who identified as White (p = 0.08) differed in recovery time, though boys who identified as Hispanic or African American (combined) recovered significantly sooner following injury (p = 0.003) (20).

A large study of 1,733 patients (79.3% White) found that youth identified as White were more likely to have prolonged recovery (symptom resolution >4 weeks post injury), compared to youth identified as Black/African American (12.1% of the sample) or Other race (8.6% of the sample; further detail about the racial identities of these youth was not reported) (22). Of note, this study also reported null findings; the racial groups did not differ in terms of proportions of youth recovered within 2 weeks, 3 weeks, or 4 weeks (22). A retrospective case-control study from a specialty concussion clinic compared 40 youth diagnosed with “post-concussion syndrome” (i.e., symptoms lasting >3 months following injury) to matched controls (matched on age and sex) who recovered within 3 weeks (21). The prolonged recovery group included 90% White and 10% Black youth, and the controls included 75.0% White and 22.5% Black youth as well as 2.5% with “unknown” race; these differences were not statistically significant (p = 0.13) (21). Of note, the prolonged recovery group contained only four youth identified as Black.

Eight studies were identified that examined the potential association between race and ethnicity and outcome from sport-related concussion among college-aged individuals (23–30). Across the eight studies there were a total of 689 young adults studied. Study settings included primary care sports medicine clinics, an outpatient hospital-based concussion program, and multiple National Collegiate Athletics Association (NCAA) institutions.

Among the studies included in this review, none were designed and carried out considering social determinants of health (SDoH) as a guiding framework. Further, no study specifically referenced or discussed social determinants of health. With that said, results of our content analysis revealed that five studies (41.7%) addressed a SDoH or closely related topic in some form or fashion (19–23). One study addressed three SDoH domains (21), one study addressed two domains (19), and the remaining three studies addressed one domain (20, 22, 23). The most commonly addressed domain was Healthcare Access and Quality (k = 3) (19, 21, 22). Two studies addressed Economic Stability (19, 20). Two other studies addressed Education Access and Quality (21, 23). One study addressed Social and Community (21). No studies addressed the Neighborhood and Built Environment domain.

No studies directly examined or studied health equity. Only one study referenced or discussed the topic of health equity. Specifically, in their discussion section, Morgan and colleagues position their findings in the context of considerable literature demonstrating childhood health disparities associated with race (21).

Four studies (33.3%) excluded participants based on demographic or health factors (19, 20, 23, 28). They were not excluded based on race. Participants were most commonly excluded for learning disabilities (23), psychiatric disorders (19, 20, 23), or a history of inpatient treatment (28), or substance use disorders (23, 28).

Five studies (41.7%) discussed future directions and research needs that implicated or were related in some form or fashion to social determinants or health equity (19, 20, 23, 26, 28). In terms of SDoH domains, social and community (19, 20, 23, 26, 28), economic stability (19, 23), and education access and quality (19, 23) were most commonly implicated in the context of future clinical and research directions.