Trends and Disparities in Inpatient Rehabilitation of Adolescents: The Effect of Demographics, Injury Characteristics, and Facility Type

Abstract

Background:

Intensive rehabilitation of adolescents occurs in general and pediatric inpatient rehabilitation facilities (IRFs), but differences in admission and outcomes by facility type have not been thoroughly investigated, particularly among persons with spinal cord injury (SCI).

Objectives:

To investigate factors related to admission to pediatric or adult IRFs among adolescents and compare the rehabilitation outcomes between facility types.

Methods:

Using a single-center retrospective cohort study design, demographic information, medical data, and rehabilitation outcomes were obtained by chart review of patients aged 15 to 18 years who received a rehabilitation medicine consultation at an urban level 1 trauma center between 2017 and 2019 (N = 96). Analysis was performed using R statistical software.

Results:

SCI was the second most common diagnosis (21% of patients) and accounted for 36% of inpatient rehabilitation admissions. SCI patients were more likely to undergo rehabilitation at the level 1 trauma versus pediatric facility (relative risk [RR] = 2.43; 95% confidence interval [CI] = 1.08–5.44) compared to traumatic brain injury patients. Admission to trauma versus pediatric IRF was also associated with Black compared to White race (RR = 2.5; CI = 1.12–5.56), violence compared to other etiologies of injury (RR = 2.0; CI = 1.10–3.77), and Medicaid compared to private insurance (RR = 2.15; CI = 1.01–4.59). Admission to pediatric IRF was associated with longer length of stay than admission to adult IRF when adjusted for diagnosis (30.86 ± 21.82 vs. 24.33 ± 18.17 days; p = .046), but Functional Independence Measures did not differ.

Conclusion:

Adolescents with SCI and those experiencing systemic disadvantages, including racism and poverty, were more likely to be admitted to trauma compared to pediatric IRF.

Introduction

Systemically marginalized populations, including racial and ethnic minority groups, have decreased access to inpatient rehabilitation (IPR) compared to dominant groups. This disparity has been consistently demonstrated in adults, including those with stroke,¹ spinal cord injury (SCI),² traumatic brain injury (TBI),^3–5 and orthopedic conditions.6,7 In contrast, data on disparities in IPR admission among pediatric patients are limited and conflicting, hampering the ability to design interventions to optimize equity in rehabilitation care for children. Some studies have shown increased discharge to IPR after acute injury among Black children compared to White children,^8,9 whereas a larger recent study has shown the opposite pattern.¹⁰ These conflicting findings may be related to differences in injury severity among these studies. The first two studies focused on moderate to severe injury and excluded patients with mild injury.^8,9 The third study included all children meeting hospital admission criteria, including those with mild injury, who represented 65% of the study population. In the subgroup of patients with severe injury, Black and White children were equally likely to discharge to IPR from acute hospitalization.¹⁰ This suggests that when a clear need for intensive rehabilitation is clinically indicated based on injury severity, there are limited differences in referral and admission patterns to IPR; however, when clinical factors are less definitive, disparities emerge.

In addition to disparities in rates of admission to IPR overall, disparities in facility type may influence outcomes. Children, particularly adolescents, may be admitted to either pediatric or general (mixed adult and pediatric) inpatient rehabilitation facilities (IRF) for IPR, but there are no guidelines or recommendations to assist clinicians in determining where adolescents should undergo rehabilitation. Only two previous studies have compared the race and ethnicity data of pediatric and general IRFs. In one study of TBI patients, Hispanic children were less likely to undergo rehabilitation at a pediatric hospital compared to both White and Black children.¹¹ The second study noted significant differences in the overall distribution of racial and ethnic groups between pediatric and general IRFs but did not further analyze the disparities by specific racial or ethnic groups.¹² Both of these studies evaluated national databases and did not analyze disparities locally or within institutions. Thus, it is possible that racial and ethnic disparities could be explained by geographic differences in demographics and their overlap with the availability of pediatric rehabilitation. It is also possible, however, that factors within individual health care facilities, such as implicit bias among health care workers and structural features of medical institutions, create disparities in rehabilitation disposition, but there are no studies investigating this possibility.

To begin to answer the question of whether disparities exist at the institutional level, we evaluated the rehabilitation disposition of adolescents admitted to Harborview Medical Center, the level 1 trauma center (LITC) affiliated with the University of Washington, which serves both adult and pediatric patients. Adolescents between 15 and 18 years of age may be admitted to the specialized level 1 trauma IRF unit within the LITC hospital or to the specialized pediatric IRF with the affiliated pediatric hospital, Seattle Children’s Hospital. Because there were no formal guidelines determining the placement of patients at these two facilities, there is the opportunity for structural and interpersonal bias to create disparities in access. We performed a retrospective chart review of adolescent patients referred to the rehabilitation medicine (PM&R) service at the LITC. We hypothesized that there would be racial or ethnic disparities in postacute care disposition between the trauma and pediatric IRFs and that these disparities could be mediated by social factors, particularly violent etiology of injury.

Methods

Participants

Patients were identified by screening the PM&R Consult Tracking Database, which documents all consultations placed to the PM&R service at the LITC since January 1, 2017. Inclusion criteria were admission to the LITC, completion of a PM&R consultation by the time of discharge, and age 15 to 18 years. Exclusion criteria were the completion of a PM&R consultation during a previous admission during the study period and age less than 15 or greater than 18 years. All records available in the current database at the time of study initiation were reviewed (2017–2020), yielding 96 unique patient records for inclusion in the analysis. See Table 1 for participant demographics and Table 2 for diagnoses.

Table 1.

Study population (N = 96)

Demographics	Mean ± SD or % (n)
Age, years	16.8 ± 0.99
Gender
Male	72.0% (69)
Female	27.0% (26)
Transgender	1.0% (1)
Race and ethnicity
White	44.8% (43)
Black/African American	15.6% (15)
Hispanic	17.7% (17)
Asian	9.4% (9)
Native Hawaiian/Pacific Islander	3.1% (3)
American Indian/Alaska Native	4.2% (4)
Multiple	4.2% (4)
Unknown/unreported	1.0% (1)
Primary language
English	93% (89)
Spanish	6.3% (6)
Arabic	1.0% (1)
Health insurance
Private	51.0% (49)
Medicaid	41.7% (40)
Tricare	4.2% (4)
Uninsured	2.0% (2)
Unknown	1.0% (1)

Table 2.

Injury and diagnosis distribution (N = 96)

Injury characteristics	% (n)
Traumatic brain injury	49% (47)
Mild	10.6% a
Mild complicated	19.1%
Moderate	6.4%
Severe	59.6%
Unknown	4.3%
Spinal cord injury	20.8% (20)
Cervical, motor complete	25.0% a
Cervical, motor incomplete	20.0%
Thoracic, motor complete	25.0%
Thoracic, motor incomplete	20.0%
Lumbar, incomplete	10.0%
Nontraumatic brain injury	8.3% (8)
Peripheral nerve or plexus	8.3% (8)
Polytrauma	8.3% (8)
Other	5.2% (5)

^aPercentages in the subcategories indicate percentages of the population within that category, not percentages of the whole population.

Data collection

Retrospective chart review of the medical records for the patients identified was conducted to gather demographic data, medical details, and acute care discharge location. Demographic data included age, gender, race and ethnicity, home zip code, insurance type, and patient’s primary language. In the medical record system, any number of racial or ethnic categories can be selected and include the following options: White/Caucasian, Black/African American, Hispanic/Latino, Asian, Native Hawaiian or Pacific Islander, American Indian or Alaska Native, other, and unknown/not reported. For statistical analysis of racial and ethnic categories, only one primary racial or ethnic category was retained per patient. If more than one racial or ethnic category was recorded, the non-White category was retained. Patients with two non-White categories selected were reported as multiracial. Racial and ethnic categories were further combined into White, Black/African American, Hispanic, and other.

Diagnostic and etiologic details were obtained from the initial PM&R consultation note. The predetermined diagnostic categories were SCI, TBI, peripheral nerve (PN) or plexus injury, amputation, and other. The predetermined etiologic categories were motor vehicle collision, violence-related injury, other trauma, medical, surgical, and psychiatric. For patients admitted to IPR at the trauma IRF within the LITC or at the pediatric IRF at the affiliated pediatric hospital, rehabilitation outcome measures were collected. Functional independence measure (FIM) scores were recorded at admission and discharge. The dates of admission and discharge and any transfers to an acute care service during IPR were recorded to determine length of stay (LOS). Days off the rehabilitation service for acute medical concerns after transfer were not included in the LOS. The discharge from IPR was recorded as home, acute care, or care facility.

Statistical analysis

All statistical analyses were performed in RStudio. Comparisons of categorical variables based on demographic groups and diagnostic categories were computed as risk ratios with 95% confidence interval indicating significance. For direct comparisons of continuous variables based on categorical factors that were amenable to separation into two groups, we utilized two-tailed Welch’s t test with unequal variance. These methods were used to describe the likelihood of types of injuries and the types of postacute care discharge locations based on demographic factors. To analyze the rehabilitation outcomes of patients, we utilized single and multiple linear regression to determine the associations between FIM change, FIM efficiency, and LOS with patient characteristics. Additional statistical tests utilized are identified in the results. The IRB was approved June 18, 2020, by the institutional review board.

Results

Etiology of injury

Motor vehicle-related trauma was the most common etiology of injury (42.7%). Violence-related trauma accounted for 19.8% of injuries. Black and Hispanic adolescents were more likely to experience a violent etiology of injury compared to White adolescents (relative risk [RR] = 4.3, 95% confidence interval [CI] = 1.4–13.2, for Black vs. White; RR = 3.79, CI = 1.22–11.79, for Hispanic vs. White). Violence-related injury was more likely to occur in individuals with Medicaid compared to private insurance (RR = 3.43, CI = 1.35–8.71). Due to the interaction between race and ethnicity with insurance type, the independence of these factors in relation to etiology of injury could not be assessed. Although median income by home zip code was lower among individuals with violence-related injury compared to other injury etiologies ($56,772 ± 22,508 vs. $62,102 ± 14,837), this was not statistically significant (p = .22, two-tailed Welch’s t test with unequal variance). Among violence-related injuries, 84.2% were gun-related, 5.3% were stabbing, and 10.5% were blunt force trauma. Most of these cases were due to assault (68.4%), followed by accidental (21%) and self-inflicted (10.5%) injuries. PN or plexus injuries were relatively frequent, representing 31.5% of violence-related injuries and accounting for 75% of patients in the study population with a primary diagnosis of PN or plexus injury.

Post-acute care rehabilitation

Among all patients included in the analysis, 45.8% (n = 44) were admitted to IPR after acute care hospitalization. An equal number of patients were discharged directly home from acute care, whereas six were discharged to a skilled nursing or long-term acute care facility (SNF/LTAC), and two died during hospitalization. There were no clear predictive factors of discharge to SNF/LTAC, but analysis was limited by the small number of patients in this group. There were no differences by race/ethnicity, etiology type (violent vs. nonviolent), or insurance (private vs. Medicaid/CHIP) in likelihood of admission to IPR compared to discharge home after acute care hospitalization. Diagnostic category did impact the probability of discharge to IPR compared to direct discharge home. Among SCI patients, 80% required IPR compared to 42.5% of TBI patients (RR = 1.77, CI = 1.22–2.57). The combination of patients with TBI and SCI accounted for 86% of those admitted to IPR.

Among patients requiring IPR after acute hospitalization, 47.7% were admitted to the pediatric IRF, 40.9% were admitted to the trauma IRF, and 11.4% were admitted to IRFs outside of the system. Within the same academic system, adolescent patients were more likely to undergo IPR at the trauma compared to pediatric facility if they were Black compared to White (RR = 2.5, CI = 1.12–5.56), had experienced violence as the etiology of injury (RR = 2.0, CI = 1.10–3.77), or had Medicaid compared to private insurance (RR = 2.15, CI = 1.01–4.59). Because of the correlations between race/ethnicity, insurance, and etiology of injury, we could not determine the independence of these associations. A primary diagnosis of SCI compared to TBI was also associated with admission to the trauma rather than pediatric facility (RR = 2.43, CI = 1.08–5.44); however, neither severity of TBI nor segmental level or completeness of SCI affected facility disposition. Median income by home zip code was lower for those remaining in the general facility ($57,790 ± 22,159) compared to the pediatric facility ($69,957 ± 2,388), but this was not statistically significant (p = .13, two-tailed Welch’s t test with unequal variance). Disposition to the pediatric or trauma IRF did not differ among the five attending physicians who performed PM&R consultations (p = .288, chi square).

Outcomes

Details of the rehabilitation course were available for patients admitted to both the trauma and pediatric facilities within the system and were compared to evaluate differences in outcomes. The overall IPR LOS for both facilities was 27.9 ± 20.2 days, and there were significant differences by facility and diagnosis. Patients with SCI had an average LOS of 37.7 ± 22.8, which was significantly longer than other groups (p = .038, two-tailed Welch’s t test with unequal variance). Patients undergoing IPR at the pediatric facility had longer LOS (30.9 ± 21.8 days) compared to patients at the trauma facility (24.3 ± 18.2). Because of the imbalance in the proportion of SCI patients between the two facilities, multiple linear regression was used to analyze the differences in LOS between facilities to account for this; both a diagnosis of SCI and admission to the pediatric rehabilitation facility were independent predictors of longer LOS in this model (p = .046 for facility type, β = 14.3, CI = 2.6–27.0; p = .049 for diagnosis of SCI, β = 20.1, CI = 3.5–36.7).

Among all patients, the overall FIM change was 33.9 ± 18.0, and the FIM efficiency was 1.77 ± 1.26. SCI patients had a lower change in FIM compared to patients with other diagnoses (29.9 ± 13.9 vs 36.1 ± 19.9), but this was not statistically significant (p = .27, Welch’s t test), and diagnosis was not associated with FIM change in a linear model. Given the longer LOS for SCI patients, FIM efficiency was significantly lower in this population compared to other diagnoses (1.10 ± 0.74 vs 2.14 ± 1.34, p = .003, Welch’s t test). Although the unadjusted FIM change was greater at the pediatric facility compared to the trauma facility (39.4 ± 20.8 vs. 27.4 ± 11.8), this was not significant when corrected for diagnosis (p = .12, multiple linear regression for facility type and diagnosis). FIM efficiency also did not differ between facilities.

Disparities in outcomes by race/ethnicity, insurance, and etiology were evaluated using multiple linear regression accounting for facility and diagnosis. LOS was significantly longer for Black patients compared to other racial and ethnic groups (42.5 vs. 20.8, p = .01) in this model; facility and diagnosis of SCI remained significant in this analysis (p = .008, β = 18.8, CI = 7.6–30.1, and p = .03, β = 20.1, CI = 4.9–35.4, respectively). Notably, there was no significant difference in the proportion of SCI diagnoses in Black compared to White patients either in the total patient sample or among the patients admitted to IPR. While differences were observed in LOS by insurance and etiology in single-variable models, these differences were not significant in the model accounting for facility and diagnosis. Both FIM change and FIM efficiency were not significantly different by race or ethnicity, insurance, or etiology when adjusted for facility and diagnosis.

Discussion

In this single-system retrospective study, we found that adolescents impacted by systemic disadvantage, including bias, poverty, and exposure to violence, were less like to undergo IPR at a specialized pediatric facility compared to a specialized trauma facility compared to their peers, despite acute hospitalization in the same LITC. This is the first study to identify IPR admission disparities between pediatric and nonpediatric IRFs occurring within individual health care systems, implying that at least some disparities are amenable to interventions within institutions. Additionally, we found that LOS was longer at the pediatric compared to trauma IRF, indicating that disparities in admission to different facilities according to race, ethnicity, and socioeconomic factors produce meaningful differences in the rehabilitation care received by systemically marginalized populations.

Our findings suggest that the disparities in admission between facility types at a national level identified in prior studies^11,12 likely reflected not only hospital-independent factors, such as geographic differences in the availability of pediatric IRFs, but also hospital-level factors, such as institutional policy and bias, which require different interventions. In contrast to hospital-independent factors, hospital-level factors can be modified by clinicians, particularly by targeting the referral and admission screening processes to improve equity in the provision of rehabilitation services among adolescents.

Violent etiology of injury and other factors caused by structural racism and economic inequality may mediate the observed racial disparities in facility-type admission patterns. Interpersonal racism and implicit bias may also play a role, though we did not observe differences in referral patterns between providers. Anecdotally, admission to the pediatric facility may be declined for adolescents who are thought to pose a potential security risk, which may include those injured in gang-related or targeted violence and those with a history of involvement with the criminal justice system. Other social factors such as lack of school enrollment and use of recreational drugs have been considered indicators of a poor fit for a pediatric facility. Decisions based on these factors, which may not be explicit in the medical record, are not formal guidelines, but knowledge of such preferences discourages physicians from referring patients to the pediatric facility. Although such considerations could be framed as technically race-neutral, the sociopolitical history of the United States ensures that Black children disproportionately experience violence and other adverse childhood events. Our findings of increased violent etiology of injury among Black and Hispanic adolescents is in accord with prior research.¹³ Prior research has demonstrated that minority children have an increased incidence of violence-related injuries compared to White children.^14–17 The disparity in incidence of violence injury rapidly increases during adolescence^14,17,18 and declines in early adulthood.¹⁹

To address disparities in admission to IPR within institutions, clinicians must be aware of the potential for implicit bias and institutional policies that create unequal care, as demonstrated in this study. Education, antiracism training, and increased representation of underrepresented groups among rehabilitation professionals are important strategies to combat health disparities, but explicit policy solutions and guidelines may be more immediately effective. Systems with both general and pediatric IRFs able to admit adolescents should consider the use of fixed age-based guidelines to determine admission to each facility type. Although such policies limit flexibility, they prevent the use of racially correlated considerations and the effects of interpersonal racism and bias, thus ensuring equity for this unique population.

In contrast to prior studies demonstrating better cognitive outcomes in children admitted to a dedicated pediatric facility,^11,12 we did not find any difference in outcomes based on FIM change between our two facilities. This may suggest that the benefits seen in prior studies were related to the needs of younger children for whom specialized pediatric therapy is more critical; however, older adolescents with SCI may benefit from specialized SCI centers, which have been associated with improved outcomes compared to nonspecialized centers.^20,21 Additionally, the longer LOS at the pediatric compared to trauma IRF is the opposite pattern observed in a prior nationwide study in which pediatric patients had longer LOS in general facilities.¹² This discrepancy could be related to possible differences between decline in IPR LOS over time between general and pediatric facilities or may reflect regional or institutional differences. Our clinical experience suggests that insurance companies tend to permit longer lengths of stay for pediatric patients. Although shorter LOS with equivalent FIM changes can be considered a measure of increased efficiency and may decrease nosocomial risks and time away from home and community, the increased LOS at the pediatric facility may provide additional time for psychological support, return to school preparation, and family training. Finally, we did not observe differences in the overall rate of IPR admission by race/ethnicity. Taken in the context of prior findings demonstrating decreased IPR admission for Black compared to White children with mild injuries,¹⁰ our findings could indicate that disparities in admission to IPR could result from decreased consultation of PM&R and therapy services for patients with mild injuries, as our study evaluated only those patients evaluated by the PM&R service whereas the prior study used a trauma database.

The main limitation of this study was the small sample size. There were only 20 adolescents with SCI available for inclusion, limiting our ability to further evaluate by level of injury or International Standards for Neurological Classification of SCI (ISNCSCI). Our results represent a single institution over a short period of time, and different patterns could be seen in other facilities and time periods. Given the sample size and correlations between race/ethnicity, insurance type, and etiology of injury, we were not able to determine the level of independence of the associations between demographic and etiologic factors and IPR facility type. Further research is necessary to characterize the mechanisms by which complex factors interact to produce disparities in IPR admission and evaluate the effectiveness of interventions.