Racial, Gender, and Neighborhood-level Disparities in Pediatric Trauma Care

Abstract

Background:

Disparities in trauma outcomes and care are well established for adults, but the extent to which similar disparities are observed in pediatric trauma patients requires further investigation. The objective of this study was to evaluate the unique contributions of social determinants (race, gender, insurance status, community distress, rurality/urbanicity) on trauma outcomes after controlling for specific injury-related risk factors.

Study Design:

All pediatric (age<18) trauma patients admitted to a single Level 1 trauma center with a statewide, largely rural, catchment area from January 2010 to December 2020 were retrospectively reviewed (n=14,398). Primary outcomes were receipt of opioids in the emergency department, post-discharge rehabilitation referrals, and mortality. Multivariate logistic regressions evaluated demographic, socioeconomic, and injury characteristics. Multilevel logistic regressions evaluated area-level indicators, which were derived from abstracted home addresses.

Results:

Analyses adjusting for demographic and injury characteristics revealed that Black children (n=6,255) had significantly lower odds (OR=0.87) of being prescribed opioid medications in the emergency department compared to White children (n=5,883). Children living in more distressed and rural communities had greater odds of receiving opioid medications. Girls had significantly lower odds (OR=0.61) of being referred for rehabilitation services than boys. Post hoc analyses revealed that Black girls had the lowest odds of receiving rehabilitation referrals compared to Black boys and White children.

Conclusion:

Results highlight the need to examine both main and interactive effects of social determinants on trauma care and outcomes. Findings reinforce and expand into the pediatric population the growing notion that traumatic injury care is not immune to disparities.

Trauma, including unintentional injury, assault, and suicide, is the leading cause of death among children in the United States [1]. An estimated 9 million children are treated for injuries each year in the emergency department (ED), with an estimated 12,000 such injuries resulting in death [2]. Following discharge, many pediatric trauma survivors will experience functional impairment [3], disruptions in their daily activities [4], and their families will experience both personal and economic strain [5]. Estimates of the financial burden of pediatric trauma vary, with more conservative figures suggesting annual direct and indirect costs of $183 billion [6]. While evidence suggests overall improvements in trauma outcomes and care over the past decade [7], not all population groups have benefitted equally. Disparities across a wide range of trauma outcomes (e.g., mortality, disability) and care (e.g., opioid prescriptions in the ED, referrals to rehabilitation services) are well established for adults [8], but significant gaps remain in our understanding of pediatric trauma disparities [9]. Social determinants of trauma disparities include both patient characteristics (e.g., race, gender) as well as the conditions in which patients live, learn, work, play, and age [10]. Disentangling the social determinants that contribute to disparities in pediatric trauma outcomes and care may help identify barriers, which in turn may lead to approaches that enhance outcomes for underserved groups.

Race has received considerable attention as a social determinant contributing to trauma disparities. Numerous studies have found that Black children have higher rates of mortality following trauma compared to their non-Hispanic White counterparts [11]. For example, a study of a large children’s hospital trauma database from 1994–2004 found that Black children had a significantly higher overall mortality rate than White children (9.6% vs. 2.8%), even after controlling for insurance status and mechanism of injury [12]. Such findings are consistent with a recent meta-analysis that showed that non-White patients, both adult and pediatric, had an 18% higher mortality rate following trauma compared to White patients [13]. Black children are also more likely to experience severe functional impairment as a result of their injuries compared to White children [14]. Additionally, some studies have reported longer ED stays for Black children [15, 16], whereas others do not support such disparities [17]. Although race plays a role in pediatric trauma disparities, the extent to which race itself drives these disparities independently of other social determinants of health remains unclear.

Social determinants associated with socioeconomic status (SES), including insurance status, have been implicated in trauma disparities. Black children and their families are more likely to be uninsured or publicly insured, with such differences believed by some to contribute, in part, to apparent racial disparities in trauma outcomes [18]. Indeed, the relationship between insurance status and trauma outcomes is well-established in adults [8]. Although less is known about the role insurance status plays in pediatric trauma outcomes, some evidence indicates that uninsured or publicly insured children face a higher mortality rate than those with private insurance [19]. Insurance status also may affect the type of trauma care received. A National Trauma Bank study of children and adults (ages 14 to 89) found that uninsured patients diagnosed with a traumatic brain injury (TBI) were less likely to undergo treatment for TBI (i.e., craniectomy/craniotomy, ventriculostomy, intracranial pressure monitor placement) and stayed in the hospital longer compared to those with insurance, even after controlling for differences in injury severity [13, 20]. The adverse effect of public insurance or lack of insurance on trauma outcomes may be more pronounced in Black compared to White children and adults, as uninsured and publicly insured Black patients exhibit higher mortality rates than uninsured White patients [13, 21]. Consistent with a role for insurance status in apparent racial disparities, some research suggests that universal health insurance mitigates racial disparities in post-discharge outcomes for children [17]. It remains unclear whether insurance status is a proxy for yet other individual characteristics (e.g. SES [22], injury severity [21]), or if it is directly linked to disparities in healthcare quality or access [9]. As such, understanding its relative importance in the context of correlated factors is key.

Place, or the characteristics of where a trauma occurs, may also be relevant to understanding trauma disparities, potentially capturing not only an individual’s SES, but also the resources they are afforded and the unique risks associated with the social environment where the trauma occurred (e.g., chronic elevated stress due to high community violence). Prior research reveals that trauma center proximity, which influences time to treatment, is a well-established determinant of trauma outcomes [23]. However, less is known about other aspects of place, including socioeconomic characteristics of the communities in which injuries occur and to which discharged patients must return for recovery. In addition, most studies evaluating community- and Census block-level factors that may contribute to health outcomes have done so in adult populations. Relatively little is known about the relationship between place and pediatric trauma outcomes and care. One measure useful for assessing potential impacts of place on trauma disparities is the Distressed Community Index (DCI), which captures the economic well-being of geopolitical areas (i.e. zip codes, city, county) using a composite of seven socioeconomic metrics [18]. Using measures such as the DCI, distressed counties have been shown to have a 52% higher mortality rate from self-harm and violence compared to non-distressed counties [18]. Residents of economically marginalized areas are also at increased risk for cardiovascular disease [24], functional disability [25], and chronic pain [25]. Moreover, distressed communities are disproportionally rural [18]. Residents of rural areas, like those of distressed communities, experience higher mortality rates following trauma [26, 27]. In a recent meta-analysis of pediatric patients presenting with a TBI, children living in more rural areas experienced reduced access to level I and II trauma centers compared to those living in more urban areas [27]. Despite promising evidence for geospatial characteristics influencing adult trauma disparities, the role of place in pediatric trauma outcomes and care remains unclear.

In summary, while disparities in adult trauma outcomes and care are well-established, the extent to which similar disparities may be observed in pediatric trauma patients requires further investigation. For example, although women are more likely to receive and to use opioid prescriptions compared to men [28], few studies have been conducted in pediatric populations [9]. In addition, the unique contribution of social determinants to pediatric trauma disparities, both at the individual-level (i.e., race, gender, insurance status) and the neighborhood-level (i.e., community distress, rurality/urbanicity), has yet to be determined [29]. The existing literature on pediatric trauma disparities is limited in its scope, with studies often focusing on a single type of injury (e.g., blunt trauma) or intent (e.g., unintentional). The present study evaluated pediatric trauma disparities in outcomes (i.e., mortality) and care (i.e., opioid prescriptions in the ED, referral to rehabilitation services) and tested the unique influences of race, gender, insurance status, and geospatial features (i.e., community distress, rurality/urbanicity) on these disparities. We hypothesized that being Black, male, uninsured, and living in more rural and distressed areas would each be associated with lower odds of being prescribed opioids in the ED, lower odds of referral to rehabilitation services, and higher odds of death.

METHODS

Data source

All pediatric patients (age < 18 years) who were admitted to the level 1 trauma center at University of Mississippi Medical Center (UMMC) between January 2010 and December 2020 were included in a retrospective review (n=14,398). Data from trauma patient encounters on demographics, injury characteristics, and insurance status were extracted directly from the UMMC institutional trauma registry. Patients’ home address data were then used to derive additional geospatial indicators of community distress and rurality. Eligibility for inclusion in the trauma registry was based on ICD-10 Codes (i.e., injuries to specific body parts [S00-S99 with 7^th character modifiers of A, B, or C only], unspecified multiple injuries [T07], injury of unspecified body region [T14], traumatic compartment syndrome [T79.A1-T79A9 with 7^th character modifier of A only]) in addition to patients meeting at least one of the following criteria: trauma team activation; hospital admission; transfer to a facility by Air Ambulance; triaged to hospital by emergency management system (EMS); transfer between acute care facilities by EMS; or death. Pediatric patients with burn injuries, late effects (≥ 30 days posttraumatic amnesia), or foreign bodies were excluded from the trauma registry. This study was approved by the institutional review board at UMMC.

Study population

This study included all pediatric trauma registry patients who self-reported their race as either Black/African American (n=6,255) or White/Caucasian (n=5,883). Pediatric patients who self-identified as either Hispanic (n=171) or other races (Native American [n=118], Asian [n=43], Pacific Islander [n=1], Other [n=382], Mixed [n=16], unknown [n=484]) were excluded to permit direct comparisons between Black and White patients. Meta-analytic reviews highlight clear trauma disparities for Black and Asian patients compared to White patients, with more inconsistent findings reported for contrasts between Hispanic and White patients [8]. The present study focused on Black-White comparisons based on both conceptual reasons (i.e., stark health disparities for Black residents in the state of Mississippi [30]) and methodological limitations (i.e., relatively small number of patients from other racial and ethnic groups). This study focused on patients with blunt and penetrating injuries.

Trauma Outcomes and Care

Primary outcomes were opioid medications prescribed in the ED, post-discharge rehabilitation referrals, and in-hospital mortality. Opioid medications (e.g., oxycodone, hydrocodone, morphine, buprenorphine, fentanyl, methadone) were coded as either 1 (one or more opioid medications were prescribed) or 0 (no opioid medications prescribed). Rehabilitation services included rehabilitation centers, home health, skilled nursing facilities, and trauma centers (1 = referral; 0 = no referral).

Predictors

Demographic data included age, gender (male, female), and race (non-Hispanic Black or White). Injury characteristics included type (i.e., blunt, penetrating), intent (i.e., unintentional, assault, self-inflicted), severity (i.e., Injury Severity Score [ISS]), vital signs obtained in the ED (i.e., heart rate, systolic/diastolic blood pressure), time spent in the ED, ventilator use, intubation, and year of admission. Socioeconomic information included insurance coverage (yes/no) and DCI score. Patient home address zip codes were used to compute DCI scores (https://eig.org/dci/interactive-map), which are a composite of 7 indicators obtained from the American Community Survey 5-year estimates (2014–2018): % of adults ages 25 and older without a high school diploma or equivalent; % of population living below the federal poverty line; % of adults ages 25–54 not working; % of housing units that are vacant; median household income as percentage of the state median household income; % change in number of employees working from 2014 to 2018; and % change in number of business establishments from 2014 to 2018 [31]. DCI scores range from 0 (no distress) to 100 (severe distress). Rurality/urbanicity for patient home address counties were computed using the 2013 Department of Agriculture Rural-Urban Continuum Codes (RUCC). RUCC codes range from 1 (metropolitan areas with a population of 1+ million) to 9 (rural areas with a population less than 2,500, not adjacent to a metropolitan area) [32]. A categorical variable reflecting time from injury to the ED (i.e., less than 1 hour, 1–6 hours, 7–12 hours, 13–24 hours, more than 24 hours, unknown) was included in sensitivity analyses.

Data analysis

Tests for unadjusted differences between non-Hispanic White and Black patients used chi-square tests for categorical variables and t-tests for continuous variables. The independent effect of each hypothesized social determinant (race, gender, insurance status, community distress, rurality/urbanicity) on pediatric trauma outcomes was assessed, using a series of models described below (as recommended in recent trauma disparities research [33]). To address hypotheses that being Black, male, and uninsured [9, 19] would be associated with lower odds of being prescribed opioids in the ED, lower odds of referral to rehabilitation services, and higher odds of death, multivariate logistic regression analyses were conducted using SPSS 27 (IBM Corp., Armonk, NY) to examine associations between patient-level predictors and three dichotomous outcomes: receipt of any opioid medications in the ED (yes/no); referral to rehabilitation services (yes/no); and mortality (yes/no). Model 1 included demographic (age, gender, race [non-Hispanic Black or White]), injury characteristics (type, intent, severity), ED vitals (heart rate, systolic and diastolic blood pressure), time in the ED, and year of admission. Model 2 added the impact of insurance status (insured/uninsured). To address hypotheses that living in more distressed and rural areas [23, 26] would be associated with lower odds of being prescribed opioids in the ED, lower odds of referral to rehabilitation services, and higher odds of death, multilevel logistic regression analyses were conducted using HLM 8 [34] to account for pediatric patients living in the same areas (i.e., zip codes, counties) and sharing the same area-level predictors. Model 3 added the area-level impact of community distress (DCI scores) to Model 2. Model 4 added the area-level impact of rurality/urbanicity (RUCC codes) to Model 2. Trauma-relevant covariates were selected based on published guidelines [35, 36]. Analysis of mortality included the full sample (N=12,138). However, analysis of rehabilitation referrals included only patients who survived (n=11,952) and analysis of ED opioid prescribing included patients who were never intubated or placed on ventilators (n=11,485). The latter restriction was based on standardized ED protocols for managing pain in cases of intubation/ventilation that are unlikely to be influenced by social determinants of health. Sensitivity analyses tested models controlling for time from injury to ED; this variable was not included in primary analyses due to substantial missing data (23%).

Similar to recent disparities research in adult [37] and pediatric trauma patients [38], there were pronounced racial differences in patient and injury characteristics (described below). Hence, post-hoc tests included race as a moderator of trauma outcomes and care. Multiple testing was addressed through the Benjamini-Hochberg false discovery rate correction, which controls for the rate of Type I errors by adjusting the p-value based on the number of significant results in a family of tests [39].

RESULTS

Patient and Injury Characteristics

Descriptive statistics for pediatric traumatic injury patients admitted to the UMMC level 1 trauma center from 2010 to 2020 are presented in Table 1. Statistics are presented for the full sample, for Non-Hispanic White and Black patients separately, and for between-group tests comparing White and Black patients (t-tests and χ² for continuous and categorical variables, respectively). These between-group tests did not adjust for the covariates included in the logistic regression analyses presented below. Non-Hispanic Black children included a higher proportion of males compared to non-Hispanic White children. Regarding injury type, Black children were less likely to experience blunt injuries and more likely to experience penetrating injuries. Regarding injury intent, Black children were less likely than White children to experience unintentional injuries and more likely to experience assault-related injuries. Black children exhibited lower ED heart rate, higher ED systolic blood pressure, and higher diastolic blood pressure compared to White children. Regarding individual- and place-based socioeconomic indicators, Black children were more likely to live in distressed communities compared to White children. Black children were also more likely than White children to reside in urban areas. Most children (60.7%) arrived at the ED between 1 and 6 hours after their injury. Sensitivity analyses including time to ED as a covariate revealed no impact on patterns of findings in subsequent logistic regression analyses. Descriptive statistics for age groups (i.e., 0 to 5 years, 6 to 12 years, 13 to 17 years) and for minor versus major injuries (based on ISS cutoff score of 15) [40] are provided in Supplemental Table 1. Penetrating injuries, assault-related injuries, and self-inflicted injuries were more common among 13 to 17 year old children; older children were also more likely to receive rehabilitation referrals and to be prescribed opioid medications in the ED. Major injuries were more likely to be assault-related or self-inflicted and to result in death or rehabilitation referrals.

Table 1.

Descriptive Statistics Summary for Non-Hispanic White and Black Pediatric Traumatic Injury Patients.

	Total (n=12,138)	Non-Hispanic White (n=5,883)	Non-Hispanic Black (n=6,255)	White vs. Black patients
	M (SD) or N (%)	M (SD) or N (%)	M (SD) or N (%)	t or χ2
Demographics
Age	8.6 (5.6)	8.6 (5.4)	8.6 (5.7)	0.1
Gender				29.0***
Female	4,412 (36.4)	2,281 (38.8)	2,131 (34.1)
Male	7,717 (63.6)	3,598 (61.2)	4,119 (65.9)
Missing	9 (0.1)	4 (0.1)	5 (0.1)
Injury Characteristics
Type				157.4***
Penetrating	1,291 (10.6)	413 (7.0)	878 (14.0)
Blunt	10,717 (88.3)	5,408 (91.9)	5,309 (84.9)
Missing	130 (1.1)	62 (1.1)	68 (1.1)
Intent				292.1***
Unintentional	10,948 (90.2)	5,585 (94.9)	5,363 (85.7)
Assault	898 (7.4)	193 (3.3)	705 (11.3)
Self-inflicted	85 (0.7)	34 (0.6)	51 (0.8)
Missing	207 (1.7)	71 (1.2)	136 (2.2)
Injury Severity (ISS)	6.8 (6.9)	6.9 (6.8)	6.7 (7.0)	1.1
Vitals in ED
Heart rate	105.7 (27.7)	106.8 (26.5)	104.7 (28.8)	4.1***
Systolic blood pressure	119.7 (17.9)	118.7 (16.5)	120.5 (19.0)	5.4***
Diastolic blood pressure	72.5 (13.8)	71.6 (13.3)	73.3 (14.2)	6.6***
Time in ED (minutes)	253.0 (246.2)	254.0 (314.2)	252.1 (157.4)	0.4
Socioeconomic/Geographic
Insurance				1.1
Insured	6,608 (54.4)	3,263 (55.5)	3,345 (53.5)
Uninsured	1,016 (8.6)	484 (8.2)	532 (8.5)
Unspecified	4,514 (37.2)	2,136 (36.3)	2,378 (38.0)
Community distress	73.9 (26.6)	65.2 (29.0)	82.2 (20.9)	36.3***
Rural-Urban Continuum	4.3 (2.4)	4.4 (2.4)	4.2 (2.3)	5.9***
Trauma Care Outcomes
Mortality (died)	186 (1.5)	80 (1.4)	106 (1.7)	2.3
Rehabilitation (referred)a	167 (1.4)	86 (1.5)	81 (1.3)	0.6
Opioid in ED (prescribed)b	3,688 (32.1)	1,861 (33.3)	1,827 (31.0)	7.2**

^***<p.001

^**p<.01.

Note. ISS = Injury Severity Score; ED = emergency department.

^asample size without deaths = 11,952

^bsample size without intubations, ventilations, or deaths = 11,485

Area Characteristics

Community distress in this sample covered almost the full range of DCI scores from 0.1 (no distress) to 100.0 (severe distress), with a mean DCI score of 73.9 (SD=26.6). DCI quintiles in this sample were as follows: prosperous (5.1%); comfortable (9.1%); mid-tier (7.4%); at risk (20.8%); distressed (57.6%). Rurality/urbanicity covered the full range of RUCC codes from 1 to 9, with a mean RUCC code of 4.3 (SD=2.4). More children resided in nonmetropolitan (51.1%) as compared to metropolitan (48.9%) counties; 10.8% resided in completely rural areas.

Are Social Determinants Associated With Opioid Prescriptions in the Emergency Department?

Results of logistic regression analyses predicting opioid prescriptions in the ED are presented in Table 2. First, potential racial and gender differences were evaluated. Adjusting for demographic and injury characteristics (Model 1), Black children had significantly lower odds of being prescribed an opioid medication in the ED than White children (OR=0.87, p=.002). Greater odds of receiving an opioid medication was associated with older age, more recent admission year, penetrating compared to blunt injuries, unintentional compared to assault injuries, higher ISS scores, higher ED heart rate, higher ED systolic and diastolic blood pressure, and longer time spent in the ED. Second, the potential influence of insurance status was evaluated. Model 2 revealed that insurance status was not significantly associated with odds of receiving an opioid medication in the ED. Third, the potential influence of community distress was evaluated. Model 3, which accounted for nesting of patients within zip codes, revealed that children living in areas with higher community distress had slightly higher odds of receiving an opioid medication (OR=1.002, p=.024). Fourth, the potential influence of rurality/urbanicity was evaluated. Model 4, which accounted for nesting of patients within counties, revealed that children living in more rural areas had slightly higher odds of receiving an opioid medication in the ED compared to children living in more urban areas (OR=1.02, p=.036).

Table 2.

Logistic Regression Models for Any Opioid Prescription in the Emergency Department

	Model 1		Model 2		Model 3		Model 4
	OR [95% CI]	p	OR [95% CI]	p	OR [95% CI]	p	OR [95% CI]	p
Demographics
Age	1.05 [1.04–1.06]	<.001	1.05 [1.03–1.06]	<.001	1.05 [1.04–1.06]	<.001	1.05 [1.04–1.06]	<.001
Female	0.95 [0.87–1.04]	.292	0.95 [0.85–1.07]	.401	0.95 [0.84–1.06]	.352	0.95 [0.85–1.06]	.379
Black	0.87 [0.80–0.95]	.002	0.84 [0.75–0.93]	.001	0.79 [0.70–0.88]	<.001	0.84 [0.75–0.93]	.001
Injury
Admission year	1.04 [1.02–1.05]	<.001	1.06 [1.05–1.08]	<.001	1.06 [1.05–1.08]	<.001	1.06 [1.04–1.08]	<.001
Type
Blunt	Reference		Reference		Reference		Reference
Penetrating	1.37 [1.18–1.59]	<.001	1.55 [1.28–1.86]	<.001	1.51 [1.25–1.82]	<.001	1.53 [1.28–1.84]	<.001
Intent
Unintentional	Reference		Reference		Reference		Reference
Assault	0.75 [0.63–0.91]	.003	0.72 [0.57–0.91]	.007	0.71 [0.56–0.91]	.005	0.72 [0.57–0.92]	.007
Self-inflicted	1.11 [0.63–1.96]	.728	0.83 [0.36–1.92]	.669	0.86 [0.37–1.98]	.721	0.85 [0.37–1.95]	.693
Injury Severity (ISS)	1.03 [1.03–1.04]	<.001	1.03 [1.02–1.04]	<.001	1.03 [1.03–1.05]	<.001	1.03 [1.02–1.04]	<.001
Vitals in ED
Heart rate	1.00 [1.00–1.01]	.009	1.00 [1.00–1.01]	.028	1.00 [1.00–1.01]	.035	1.00 [1.00–1.01]	.026
SBP	1.01 [1.01–1.01]	<.001	1.01 [1.01–1.02]	<.001	1.01 [1.01–1.01]	<.001	1.01 [1.01–1.02]	<.001
DBP	1.01 [1.01–1.02]	<.001	1.01 [1.01–1.02]	<.001	1.01 [1.01–1.02]	<.001	1.01 [1.01–1.02]	<.001
Time in ED (mins)	1.00 [1.00–1.00]	.009	1.00 [1.00–1.00]	.008	1.00 [1.00–1.00]	.005	1.00 [1.00–1.00]	.009
Socioeconomic
Insured			0.97 [0.83–1.14]	.718	0.95 [0.81–1.11]	.504	0.96 [0.82–1.12]	.582
Geographic
Community distress					1.00 [1.00–1.01]	.024
Rural-Urban							1.02 [1.00–1.05]	.036

Note. Model 1 = Demographic and Injury Characteristics; Model 2 = Insurance Status; Model 3 = Community Distress; Model 4 = Rurality/Urbanicity; OR = odds ratio; CI = confidence interval; ISS = Injury Severity Score; ED = emergency department; SBP = systolic blood pressure; DBP = diastolic blood pressure.

Moderation by Race.

Post-hoc analysis simultaneously tested for moderation by race of the effects of age, gender, and injury type effects on opioid prescribing using a multivariate logistic regression model. Race did not moderate age (b=.003, SE=.01, p=.779), gender (b=.05, SE=.12, p=.672), or injury type (b=.08, SE=.19, p=.662) effects on odds of being prescribed an opioid in the ED.

Are Social Determinants Associated With Post-Discharge Rehabilitation Referrals?

Results of logistic regression analyses predicting post-discharge rehabilitation referral are presented in Table 3. First, potential racial and gender differences were evaluated. Adjusting for demographic and injury characteristics (Model 1), girls had significantly lower odds than boys of being referred for rehabilitation services (OR=0.61, p=.021). Greater odds of rehabilitation referral was associated with older age, more recent admissions, penetrating compared to blunt injuries, self-inflicted compared to unintentional injuries, and higher injury severity. Race was not independently associated with odds of receiving a rehabilitation referral. Second, the potential influence of insurance status was evaluated. Model 2 further revealed that insurance status was not significantly associated with odds of receiving a rehabilitation referral. Finally, the potential influence of community distress and rurality/urbanicity were evaluated. Models 3 and 4, which accounted for nesting of patients within zip codes and counties, respectively, revealed that neither community distress nor rurality/urbanicity were independently associated with odds of death (p’s>0.12).

Table 3.

Logistic Regression Models for Post-Discharge Rehabilitation Referral

	Model 1		Model 2		Model 3		Model 4
	OR [95% CI]	p	OR [95% CI]	p	OR [95% CI]	p	OR [95% CI]	p
Demographics
Age	1.08 [1.04–1.13]	<.001	1.10 [1.04–1.16]	.001	1.10 [1.05–1.17]	<.001	1.10 [1.04–1.16]	.001
Female	0.61 [0.40–0.93]	.021	0.54 [0.32–0.91]	.019	0.58 [0.35–0.97]	.038	0.58 [0.35–0.97]	.036
Black	0.89 [0.61–1.30]	.554	0.92 [0.58–1.44]	.704	0.99 [0.62–1.59]	.968	0.89 [0.57–1.40]	.608
Injury
Admission year	1.07 [1.01–1.13]	.022	1.08 [1.02–1.16]	.016	1.08 [1.01–1.15]	.027	1.08 [1.01–1.15]	.017
Type
Blunt	Reference		Reference		Reference		Reference
Penetrating	1.89 [1.02–3.48]	.042	1.25 [0.58–2.71]	.569	1.34 [0.63–2.89]	.443	1.34 [0.63–2.85]	.443
Intent
Unintentional	Reference		Reference		Reference		Reference
Assault	0.65 [0.31–1.37]	.261	0.75 [0.31–1.85]	.527	0.77 [0.32–1.87]	.562	0.75 [0.31–1.84]	.534
Self-inflicted	5.55 [1.85–16.67]	.002	10.56 [2.91–38.36]	<.001	9.70 [2.66–35.42]	.001	10.40 [2.89–37.45]	<.001
Injury Severity (ISS)	1.18 [1.16–1.20]	<.001	1.17 [1.15–1.20]	<.001	1.17 [1.15–1.19]	<.001	1.17 [1.15–1.19]	<.001
Vitals in ED
Heart rate	1.00 [0.99–1.01]	.660	1.00 [0.99–1.01]	.780	1.00 [0.99–1.01]	.716	1.00 [0.99–1.01]	.857
SBP	0.99 [0.99–1.01]	.824	0.99 [0.98–1.01]	.358	0.99 [0.98–1.01]	.274	0.99 [0.98–1.01]	.299
DBP	1.01 [0.99–1.03]	.165	1.02 [1.00–1.03]	.053	1.02 [1.00–1.04]	.047	1.02 [1.00–1.03]	.049
Time in ED (mins)	0.99 [0.99–1.00]	.193	0.99 [0.99–1.00]	.036	0.99 [0.99–1.00]	.052	0.99 [0.99–1.00]	.057
Socioeconomic
Insured			1.94 [0.85–4.43]	.115	1.93 [0.85–4.39]	.117	1.94 [0.86–4.42]	.112
Geographic
Community distress					0.99 [0.99–1.00]	.124
Rural-Urban							1.01 [0.92–1.11]	.791

Note. Model 1 = Demographic and Injury Characteristics; Model 2 = Insurance Status; Model 3 = Community Distress; Model 4 = Rurality/Urbanicity; OR = odds ratio; CI = confidence interval; ISS = Injury Severity Score; ED = emergency department; SBP = systolic blood pressure; DBP = diastolic blood pressure.

Moderation by Race.

Post-hoc analysis simultaneously tested for moderation by race of the effects of age, gender, and injury type effects on rehabilitation referrals using a multivariate logistic model. Results revealed a significant race X gender interaction (b=−1.30, SE=.55, p=.018), with Black girls having lower odds of receiving referrals than Black boys, White boys, and White girls (Figure 1). Race did not significantly moderate the effects of age (b=−.08, SE=.05, p=.078) or injury type (b=−.09, SE=.76, p=.904) on odds of rehabilitation referrals.

Figure 1.

Probability of Receiving a Referral for Rehabilitation Services for Pediatric Trauma Patients According to Race and Gender

Are Social Determinants Associated With Mortality?

Results of logistic regression analyses predicting mortality are presented in Table 4. First, potential racial and gender differences were evaluated. Adjusting for demographic and injury characteristics (Model 1), girls had significantly higher odds of dying from injuries than boys (OR=1.61, p=.045). Race was not significantly associated with odds of death. Greater odds of death was associated with more recent admission year, assault compared to unintentional injuries, self-inflicted compared to unintentional injuries, higher ISS scores, lower heart rate, lower systolic blood pressure, higher diastolic blood pressure, and less time spent in the ED. Second, the potential influence of insurance status was evaluated. Model 2 revealed that insurance status was not significantly associated with odds of death. Gender was no longer associated with odds of death after controlling for insurance status. Finally, the potential influence of community distress and rurality/urbanicity were evaluated. Models 3 and 4, which accounted for nesting of patients within zip codes and counties, respectively, revealed that neither community distress nor rurality/urbanicity were independently associated with odds of death (p’s>0.58).

Table 4.

Logistic Regression Models for Mortality

	Model 1		Model 2		Model 3		Model 4
	OR [95% CI]	p	OR [95% CI]	p	OR [95% CI]	p	OR [95% CI]	p
Demographics
Age	0.98 [0.94–1.03]	.430	0.97 [0.91–1.02]	.200	0.97 [0.92–1.03]	.293	0.96 [0.91–1.01]	.130
Female	1.61 [1.01–2.56]	.045	1.53 [0.88–2.65]	.128	1.46 [0.82–2.59]	.201	1.45 [0.87–2.42]	.154
Black	0.76 [0.48–1.22]	.259	0.67 [0.38–1.18]	.163	0.73 [0.40–1.35]	.319	0.67 [0.40–1.15]	.147
Injury
Admission year	1.09 [1.02–1.17]	.009	1.13 [1.05–1.22]	.002	1.11 [1.03–1.21]	.009	1.11 [1.04–1.20]	.003
Type
Blunt	Reference		Reference		Reference		Reference
Penetrating	1.18 [0.60–2.34]	.634	1.31 [0.57–2.99]	.522	1.35 [0.57–3.17]	.493	1.67 [0.79–3.53]	.182
Intent
Unintentional	Reference		Reference		Reference		Reference
Assault	4.65 [2.49–8.71]	<.001	5.98 [2.92–12.25]	<.001	5.66 [2.64–12.11]	<.001	5.89 [3.01–11.49]	<.001
Self-inflicted	21.02 [7.43–59.46]	<.001	20.08 [5.44–74.03]	<.001	18.86 [4.73–75.28]	<.001	20.11 [5.87–68.84]	<.001
Injury Severity (ISS)	1.21 [1.18–1.23]	<.001	1.20 [1.18–1.23]	<.001	1.21 [1.18–1.24]	<.001	1.21 [1.18–1.24]	<.001
Vitals in ED
Heart rate	0.98 [0.97–0.99]	<.001	0.98 [0.97–0.98]	<.001	0.98 [0.97–0.99]	<.001	0.98 [0.97–0.98]	<.001
SBP	0.95 [0.94–0.97]	<.001	0.96 [0.95–0.98]	<.001	0.96 [0.94–0.98]	<.001	0.96 [0.94–0.98]	<.001
DBP	1.03 [1.01–1.05]	.005	1.03 [1.00–1.05]	.023	1.03 [1.00–1.05]	.029	1.03 [1.00–1.05]	.018
Time in ED (mins)	0.99 [0.99–0.99]	<.001	0.99 [0.99–0.99]	<.001	0.99 [0.99–0.99]	<.001	0.99 [0.99–0.99]	<.001
Socioeconomic
Insured			0.61 [0.29–1.28]	.189	0.67 [0.30–1.52]	.344	0.63 [0.31–1.27]	.200
Geographic
Community distress					0.99 [0.99–1.01]	.585
Rural-Urban							0.98 [0.89–1.09]	.753

Note. Model 1 = Demographic and Injury Characteristics; Model 2 = Insurance Status; Model 3 = Community Distress; Model 4 = Rurality/Urbanicity; OR = odds ratio; CI = confidence interval; ISS = Injury Severity Score; ED = emergency department; SBP = systolic blood pressure; DBP = diastolic blood pressure.

Moderation by Race.

Post-hoc analysis using multivariate logistic regression simultaneously tested for moderation by race of the effects of age, gender, and injury type effects on mortality risk. Results revealed that race did not significantly moderate the effects of age (b=−.04, SE=.05, p=.437), gender (b=−.27, SE=.56, p=.633), or injury type (b=−.39, SE=.76, p=.611) on odds of death.

DISCUSSION

The present study sought to disentangle the independent effects of race, gender, and insurance status on trauma care (i.e., opioid prescriptions in the emergency department [ED], post-discharge referrals to rehabilitation services) and outcomes (i.e., mortality) for pediatric patients, while carefully adjusting for the potentially confounding effects of traumatic injury characteristics [36]. Prior research suggests that associations between specific individual-level factors (e.g., race) and trauma disparities (e.g., mortality) disappear when accounting for neighborhood poverty levels [33]; nevertheless, neighborhood-level social determinants are not routinely included in studies of traumatic injury outcomes and care [35]. Hence, the impact of established (i.e., rurality/urbanicity) and often-overlooked (i.e., community distress) area-level social determinants of health [27, 41] were evaluated as predictors of pediatric trauma disparities.

Prior research suggests that children presenting to the ED are less likely to receive adequate pain management, as evident in lower rates of opioid prescribing in the ED and at discharge as well as underdosing compared to adults [42, 43]. Specifically, Black children may be less likely than White children to receive analgesic medication [44]. Consistent with our hypothesis, Black children had lower odds of receiving opioid medications in the ED compared to White children. This difference remained significant after careful adjustment for potential confounds, suggesting that racial disparities in pediatric opioid prescriptions in the ED could not be attributed to differences in traumatic injury characteristics, insurance status, levels of community distress, or rural/urban residence. This finding adds to prior work in adults showing that Black patients were less likely than White patients to received pain medications during emergency medical services – despite Black patients reporting higher average pain scores than White patients [45]. Moreover, a recent meta-analytic review found that Black patients were less likely to receive analgesia for acute pain in ED settings [46]. In contrast, a large study of children with long bone fractures (n=1,191) found no significant differences between Black and White patients in odds of receiving opioid analgesic medications in the ED after adjusting for other risk factors [47]. The latter study evaluated analgesic prescribing practices across the United States using the National Hospital Ambulatory Medical Care Survey (NHAMCS) and found regional differences, such that children in the South had greater odds of receiving analgesics compared to those in the Northeast; the authors speculate that racial disparities “may be more or less pronounced in certain regions, but because of insufficient numbers, an interaction term could not be introduced into our model” [47]. Thus, one explanation for the discrepancy between studies is that the present study was conducted in a level 1 trauma center in a Southern state in which racial disparities in analgesic prescribing may be more prevalent. Other studies conducted in the Southeastern United States have reported inadequate pain management in ED settings for younger patients and Black patients [48].

Traumatic injury patients who receive rehabilitation services following discharge exhibit improved functional outcomes [49]. However, a variety of individual- (e.g., insurance status, race/ethnicity) and neighborhood-level (e.g., urban hospital settings) factors influence likelihood of receiving referrals [49, 50]. Contrary to our hypothesis, girls had lower odds than boys of being referred for post-discharge rehabilitation services, even after controlling for traumatic injury characteristics, insurance status, community distress levels, and rurality/urbanicity. Race moderated this effect, such that Black girls had significantly lower odds of receiving rehabilitation referrals than Black boys and White children. This pattern of findings is consistent with research suggesting that the joint effects of female gender and Black race exert a multiplicative effect on functional health limitations [51]. The present study extends prior work by highlighting the need to evaluate the interactive effect of gender and race on trauma care outcomes. We need to understand why Black girls had lower odds of being referred for post-discharge rehabilitation services, which improve functional outcomes [49], so that pediatric trauma disparities can be eliminated.

Studies of pediatric trauma outcomes suggest that Black, Hispanic, male, and uninsured patients are more likely to die from their injuries than White, female, and insured patients [9, 13]. The present study examined the extent to which each social determinant was independently associated with odds of death after adjusting for the effects of other social determinants. Girls had greater odds of dying from their injuries than boys after controlling for demographic and injury characteristics. However, gender differences in odds of death were nonsignificant after further adjusting for insurance status. This finding is consistent with one study showing that gender differences in pediatric trauma mortality were nonsignificant in models that controlled for insurance status [52]. Importantly, and in contrast to our recent study of adult injury patients at the same level 1 trauma center [37], we did not observe an independent effect of insurance status on mortality risk for pediatric patients. This discrepancy may be due to a higher proportion of insured pediatric patients (86.7% excluding those with missing data) compared to insured adult patients (56.3% excluding those with missing data). In Mississippi, the Children’s Health Insurance Program provides health coverage for uninsured children who are not eligible for Medicaid. Another possibility is that insurance status served as a proxy for other mortality-relevant individual characteristics that differed between boys and girls but were not directly assessed in the present study (e.g., SES [22]).

Morality risk increases as distance from injury location to trauma center increases [53]. However, other neighborhood-level characteristics such as neighborhood SES have received less attention as potential determinants of pediatric trauma care and outcomes. In contrast to our hypotheses, pediatric patients living in more distressed and rural communities were more likely to be prescribed opioid medications in the ED, regardless of demographic characteristics, injury type/severity, and insurance status. These results should be interpreted cautiously until replicated in other studies. More research is needed to evaluate whether and to what extent associations between neighborhood indices and opioid prescribing practices could be explained by differential exposure to traumatic injury types. Neither community distress nor rurality/urbanicity were associated with rehabilitation referrals or mortality risk. Prior work suggests that pediatric trauma patients living in lower SES communities are more likely to die from their injuries; however, this association is explained by a higher incidence of fatal mechanisms of injury (e.g., firearm injuries) [22]. In the present study, controlling for multiple measures of injury severity (ISS scores, ED vitals) may explain the lack of an association between community distress and mortality risk.

The present study adopted a robust risk adjustment methodology to disentangle the unique contributions of social determinants on pediatric trauma care and outcomes. However, there were important limitations that provide direction for future research. First, the trauma registry database lacked geocoded data on traumatic injury location and residential addresses for patients. Hence, we were unable to assess the effect of trauma center proximity on trauma outcomes, nor were we able to evaluate neighborhood SES with greater spatial resolution. Second, important and potentially confounding variables (e.g., acute pain levels) likely to influence trauma care (e.g., opioid prescribing) were omitted from the trauma registry database and therefore unavailable for analyses. Third, family-level SES indicators were not available; hence, we were unable to evaluate the role of factors known to influence trauma care and outcomes such as household income [22]. Finally, as with other large sample studies of traumatic patients, interpretation of the present findings should focus on effect sizes given the p-value problem [54].

CONCLUSIONS

Despite the long-held view that traumatic injury care is “immune to disparities” [8] due to the existence of standardized protocols [9], the past two decades have yielded abundant evidence of trauma disparities across the continuum of care [10]. The present study addressed key gaps in our understanding of the social determinants of trauma care and outcomes [9], including gender and neighborhood-level disparities. In contrast to our recent work showing that insured adult trauma patients were less likely to die and more likely to receive rehabilitation referrals than those without insurance [37], insurance status was not associated with trauma care and outcomes for pediatric trauma patients. Moreover, whereas Black pediatric trauma patients were less likely to be prescribed opioid medications and Black girls were the least likely to receive rehabilitation referrals in the present study, our prior work from the same level 1 trauma center suggested that race was not associated with either of these outcomes among adult trauma patients [37]. Together, these findings indicate that social determinants of trauma care and outcomes differ for pediatric and adult patients and may require unique interventions (e.g., leveraging trauma quality improvement infrastructure) for these populations to eliminate disparities.