Abstract
Objective:
To examine risk factors associated with homeboundness 1-year following traumatic brain injury (TBI) and to explore associations between homebound status and risk of future mortality and nursing home entry.
Design:
Secondary analysis of a longitudinal prospective cohort study
Setting:
TBI Model Systems (TBIMS) Centers
Participants:
Community-dwelling TBIMS participants (n=6,595) who sustained moderate-severe TBI between 2006-2016, and resided in a private residence 1-year post-injury.
Interventions:
N/A
Main Outcome Measures:
Homebound status (leaving home ≤1-2 days per week), 5-year mortality, and 2- or 5-year nursing home entry.
Results:
In our sample, 14.2% of individuals were homebound 1-year post-injury, including 2% who never left home. Older age, having less than a Bachelor’s degree, Medicaid insurance, living in the Northeast or Midwest, dependence on others or special services for transportation, unemployment or retirement, and needing assistance for locomotion, bladder management, and social interactions at 1-year post-injury were associated with being homebound. After adjustment for potential confounders and an inverse probability weight for nonrandom attrition bias, being homebound was associated with a 1.69-times (95% CI: 1.35-2.11) greater risk of five-year mortality, and a non-significant but trending association with nursing home entry by 5 years post-injury (RR=1.90, 95% CI: 0.94, 3.87). Associations between homeboundness and mortality were consistent by age subgroup (+/− 65 years).
Conclusions:
The negative long-term health outcomes among persons with TBI who rarely leave home warrants the need to re-evaluate home discharge as unequivocally positive. The identified risk factors for homebound status, and its associated negative long-term outcomes, should be considered when preparing patients and their families for discharge from acute and post-acute rehabilitation care settings. Addressing modifiable risk factors for homeboundness, such as accessible public transportation options and home care to address mobility, could be targets for individual referrals and policy intervention.
Keywords: Traumatic brain injury, rehabilitation, homebound, home discharge, discharge planning, post-acute, social isolation
Introduction:
Global public health surveillance data indicates the incidence and prevalence of traumatic brain injuries (TBI) has risen over the last three decades.1 For individuals who sustain incident TBI that requires inpatient care, discharging home is cost-saving, and generally perceived to carry a positive connotation for patients, families, providers, and insurance carriers.2-7 However, TBI may lead to chronic physical, cognitive, behavioral, and psychological challenges8-10 that complicates interactions with peers and engagement in activities outside the home, like driving, grocery shopping, or employment. As such, despite returning home (versus an institution) after inpatient care, some individuals with TBI may risk becoming homebound, and suffer downstream from the associated negative consequences.
Being homebound is defined as living in the community but never or rarely leaving the home.11 Homeboundness is associated in the general population with an elevated risk for disease burden and functional limitations, decreased access to medical services, and an elevated risk of mortality.11,12 Individuals who are homebound also experience social isolation and loneliness,13 which carries negative implications for a positive recovery trajectory, maintaining quality of life, avoiding nursing home entry, and delaying mortality.14-16 Social engagement and community participation are important goals for many people with TBI.17 Studies in older adults suggests that homeboundness threatens these goals and contributes to physical and mental health deterioration,18,19 which can in turn perpetuate social isolation.
Factors leading to homeboundness among individuals discharged home with moderate-severe TBI remain unclear, and the long-term impact of homeboundness on risk for mortality and entry into a nursing home within the TBI population are poorly understood. The goals of this study were twofold. First, we aimed to create a parsimonious prognostic model of risk factors predicting homeboundness among community-dwelling adults 1-year post-TBI. Second, we evaluated the associations between being homebound 1-year post-injury for the risk of nursing home entry and mortality up to 5 years post-injury. Clarifying details on who is at risk for becoming homebound may yield highly specific and tangible targets for intervention that may allow rehabilitation professionals and care teams to help prevent homeboundness and its potential negative consequences.
Methods:
Design
We conducted a secondary analysis using data from the TBI Model Systems (TBIMS) National Database. This prospective cohort study evaluates persons with TBI who require acute hospitalization and inpatient rehabilitation for TBI, enrolling participants during inpatient rehabilitation and following them at 1 year, 2 years, 5 years, and every subsequent 5 years post- TBI. The TBIMS National Database contains data from 19 centers across the United States, and includes over 19,000 enrolled participants in total. In follow-up interviews, participants or their proxies complete multi-dimensional self-report measures of cognition, mood, physical health, functional deficits, and community participation. All participating TBIMS centers received approval from their local institutional review boards for primary data collection.
Participants
Eligible participants for the TBIMS National Database must have sustained a moderate-severe TBI, as defined by either: post-traumatic amnesia greater than 24 hours; loss of consciousness greater than 30 minutes; Glasgow Coma Scale in the emergency department (ED) <13; or any intracranial neuroimaging abnormalities from computed tomography imaging. Participants were admitted to an acute hospital ≤72 hours of injury and were age ≥16 years upon injury. For the present study, we included participants injured prior to 2016 to allow sufficient accrual of 5-year outcome data (i.e., nursing home entry or death). There were n=6,595 participants enrolled between August 2006-July 2016. All 1-year post-injury assessments of homebound status were included before the COVID-19 pandemic (when time spent at-home was inflated), though 5-year follow-up data on nursing home entry and mortality was included up through January 2022. Because of our interest in community-dwelling individuals, we restricted our sample to participants living in a community-based private residence 1-year post-injury. We have provided a flow diagram showing the derivation of our analytic sample in Figure 1.
Figure 1. CONSORT Study Flow Diagram, by Aim.
The analytic sample for Aims 1 and 2 are described herein.
Measures
Homeboundness
Homeboundness at 1-year post-injury was characterized using the following item from the Participation Assessment with Recombined Tools-Objective (PART-O) measure20: “In a typical week, how many days do you get out of your house and go somewhere? It could be anywhere. It doesn’t have to be any place “special.” We categorized participants as homebound if they reported leaving their home “never” or “1 to 2 days” in a typical week, consistent with other work in the field.11 We considered participants endorsing other valid responses (leaving home 3-7 days/week) as non-homebound.
Riskfactors and covariates
We used a data-driven approach to evaluate 39 candidate variables that may predict 1-year homebound status. Demographic variables included age at injury, sex, race/ethnicity (White, Black, Hispanic, Other), education (less than versus greater than a Bachelor’s degree), and year of follow-up. Clinical/injury variables included acute inpatient rehabilitation length of stay, acute inpatient rehabilitation payer source (i.e., private insurance, Medicare, Medicaid, No fault auto/worker’s compensation, other), history of prior TBI, injury mechanism, acute cranial surgery, Glasgow Coma Scale, post-traumatic amnesia duration, and time to follow commands. We categorized whether participants were fully independent or not (i.e., item level score of 7 vs. 1-6) on each of the 19 items of the Functional Independence Measure (FIM) measured at year-1. We also considered pre-injury health (i.e., Charlson Comorbidity Scores calculated using acute hospital International Classification of Disease (ICD) 9th and 10th edition Diagnoses codes). We used an item regarding difficulty going outside the home alone in the 6-months prior to injury to describe pre-injury limitations in getting out of the house. We considered personal factors measured contemporaneously with homeboundness at 1-year post-injury: living arrangement (lives alone, lives with spouse/significant other, lives with parents/siblings/other family, lives with other people not specified), employment status, and primary mode of transportation. Finally, we evaluated urbanicity of residence (urban, rural, suburban classification at the zip code level21) and U.S. Census geographic region (South, Northeast, Midwest, West).
Mortality and Nursing Home Entry (Aim 2)
We evaluated five-year survival status following the 1-year post-injury interview (e.g., when homebound status was determined). Date of death was documented using the Social Security Death Index. Individuals who died within five years were followed until their date of death, and for the mortality models all others were censored either: 1) at their last known followup date if less than five years, or 2) at five years if their last known follow-up date was greater than five years. We excluded participants who had no follow-ups after their year-1 interviews, or who had an unknown date of death, because we could not confirm their censorship date for survival analyses.
Since the year-1 homebound status was evaluated among persons who were community-dwelling, we next evaluated whether survivors transitioned from community-living to long-term residence in a nursing home or subacute care at either year-2 or year-5. We excluded individuals in this model if they were deceased, lost, or incarcerated after their year-1 followup.
Statistical analysis
To create a parsimonious, prognostic model of homeboundness at 1-year post-injury, we employed a data-driven approach for variable selection because of limited prior literature studying predictors of homeboundness in TBI populations. Specifically, we used sample-splitting with cross-validation per the following steps. First, we split our dataset in half into independent training and testing sets. In our training set, we used an adaptive least absolute shrinkage and selection operator (LASSO) model. This regression shrinkage methodology facilitates variable selection by constraining the sum of squared errors of the coefficients and shrinking all predictors with unstable estimates to zero, effectively dropping these variables from the model.22-24 We inputted 39 candidate variables (see Supplemental Table 1) to our LASSO model and used 10-fold cross-validation to inform selection of our constraining term (λ). Individuals with non-missing data on candidate variables were included in the LASSO algorithm; we included the item-level variable missingness in Supplemental Table 1. For variable selection of categorical variables, we retained or excluded these variables together for evaluation in the testing sample prognostic model. Furthermore, the variables that remained in the LASSO model were then used for prognostic modeling in the independent testing set. We used a multivariable modified Poisson regression model with robust standard errors to determine associations between selected predictor variables and risk of homeboundness at 1-year post-injury in the testing set. This was done because estimates from a LASSO regression model are systematically shrunk and not in and of themselves interpretable.25
We then conducted a survival analysis evaluating the association between homeboundness 1-year post-injury and risk for mortality over the next five years. We censored participants as detailed above. We plotted a Kaplan Meier curve and the associated log-rank test, and ran a series of Cox proportional hazards regression models, inputting different covariates at each successive step to evaluate their impact on our homeboundness estimate. We tested survival model assumptions of proportional hazards using the Supremum Test for Proportional Hazards Assumption, and evaluated interactions between homebound*time to test whether the effect of homeboundness on mortality was proportional over time. Because homeboundness and risk of mortality are commonly linked to older adult populations,13,19 we conducted a sensitivity subgroup analysis, stratifying mortality models by age, ≤65 and >65 years. For the nursing home entry model, we used a Poisson regression model with robust standard errors. For both outcomes, we evaluated five successive nested models based on confounders that were conceptually selected a priori based on the extant literature of known predictors of either homeboundness, outcomes of interest,26 or both (Model 1: unadjusted; Model 2: age only; Model 3: + sex; Model 4: + race ethnicity and education; Model 5: + FIM Motor at inpatient rehabilitation discharge, length of stay of inpatient rehabilitation, and Charlson Comorbidity Index).
To empirically account for potential selection bias caused by nonrandom attrition between Aim 1 models (when the primary predictor, homeboundness, was assessed) and Aim 2 mortality and nursing home models, we derived stabilized inverse probability weights (IPW).27 We derived weights based on predicting inclusion into the Aim 2 models using logistic regression (separate weights were calculated for mortality and nursing home models) as a function of baseline demographic and clinical characteristics. Our calculated stabilized IPW were applied to all Aim 2 mortality and nursing home models.
Results:
Prevalence of Homeboundness at 1-year Post-injury
We present descriptive characteristics of the sample (n=6,595) by homebound status in Table 1. The prevalence of homeboundness at 1-year post-injury was 14.2% (935/6,595). Of these, 13.8% (129/935) were completely homebound (i.e., never leave the home in a typical week) and 86.2% (806/935) were mostly homebound (i.e., leaving the home 1-2 days/week).
Table 1.
Characteristics of the sample by homebound status at year 1 (n=6,595)
| Homebound at 1 - year Post-injury (n=935) |
Not Homebound at 1- year Post-injury (n=5,660) |
|
|---|---|---|
| Type of homeboundness, n (column %) | ||
| Completely homebound | 129 (13.8%) | n/a |
| Mostly homebound | 806 (86.2%) | n/a |
| Demographic and personal characteristics | ||
| Age at injury, mean (SD) | 49.6 (20.6) | 41.4 (19.2) |
| Age at injury category, n (column %) | ||
| 16-29 | 196 (21.0%) | 2,117 (37.5%) |
| 30-49 | 301 (32.2%) | 1,588 (28.1%) |
| 50-64 | 201 (21.5%) | 1,151 (20.4%) |
| 65+ | 237 (25.4%) | 796 (14.1%) |
| Year of follow-up interview, n (column %) | ||
| 2007-2010 | 317 (34.0%) | 1,827 (32.3%) |
| 2011-2014 | 242 (25.9%) | 1,561 (27.6%) |
| 2015-2017 | 374 (40.1%) | 2,271 (40.1%) |
| Sex, n (column %) | ||
| Male | 618 (66.1%) | 4,188 (74.0%) |
| Female | 317 (33.9%) | 1,469 (26.0%) |
| Race/ethnicity, n (column %) | ||
| White | 555 (59.4%) | 3,922 (69.3%) |
| Black | 196 (21.0%) | 801 (14.2%) |
| Hispanic | 130 (13.9%) | 690 (12.2%) |
| Other | 54 (5.8%) | 247 (4.4%) |
| Education, n (column %) | ||
| Less than Bachelor’s degree | 813 (87.6%) | 4,488 (79.8%) |
| Bachelor’s degree or greater | 115 (12.4%) | 1,137 (20.2%) |
| Clinical and injury characteristics | ||
| Acute care length of stay, mean (SD) | 22.0 (19.7) | 20.1 (16.7) |
| Inpatient rehabilitation care length of stay, mean (SD) | 27.7 (25.6) | 23.5 (19.7) |
| Inpatient rehabilitation primary payer source, n (row %) | ||
| Private insurance | 249 (26.7%) | 2,672 (47.3%) |
| Medicare | 242 (26.0%) | 747 (13.2%) |
| Medicaid | 245 (26.3%) | 1,089 (19.3%) |
| No fault auto/worker’s compensation | 80 (8.6%) | 575 (10.2%) |
| Other | 116 (12.5%) | 564 (10.0%) |
| History of prior TBI, n (column %) | 177 (20.3%) | 1,611 (29.3%) |
| Index TBI mechanism, n (row %) | ||
| Vehicular | 369 (39.5%) | 2,989 (52.9%) |
| Fall | 358 (38.3%) | 1,549 (27.4%) |
| Violence | 103 (11.0%) | 485 (8.6%) |
| Pedestrian | 76 (9.1%) | 397 (7.0%) |
| Other | 28 (3.0%) | 226 (4.0%) |
| Required cranial surgery, n (column %) | 257 (27.6%) | 1,372 (24.4%) |
| Glasgow Coma Scale (GCS) score, Median (IQR) | 14 (9-15) | 13 (8-15) |
| Duration of post-traumatic amnesia (PTA) in days, Median (IQR) | 24 (8-47) | 21 (8-37) |
| Time to follow motor commands in days, Median (IQR) | 2 (0.5-11) | 2 (0.5-9) |
| Pre-index limitations in getting out of house, n (column %) | 91 (9.8%) | 135 (2.4%) |
| Charlson Comorbidity Index, Median (IQR) | 0.7 (0-1) | 0.5 (0-1) |
| Contemporaneous characteristics at 1 year follow-up | ||
| Living arrangement, n (column %) | ||
| Alone | 363 (38.9%) | 2,287 (40.5%) |
| Lives with spouse or significant other | 363 (38.9%) | 2,287 (40.5%) |
| Lives with parents, siblings, other family | 404 (43.3%) | 2,305 (40.8%) |
| Other not specified | 44 (4.7%) | 305 (5.4%) |
| Current employment status, n (column %) | ||
| Student/Employed | 54 (5.8%) | 2,191 (38.8%) |
| Retired | 576 (62.1%) | 1,813 (32.1%) |
| Unemployed/Volunteer/Homemaker/Other | 298 (32.1%) | 1,645 (29.1%) |
| Current primary mode of motorized transportation, n (column %) | ||
| Drives vehicle | 114 (12.3%) | 2,615 (46.3%) |
| Rides with someone else | 679 (73.1%) | 2,324 (41.1%) |
| Public transit | 67 (7.2%) | 563 (10.0%) |
| Special bus/van service | 52 (5.6%) | 90 (1.6%) |
| No motor transit | 17 (1.8%) | 60 (1.1%) |
| Geographic characteristics of residence at 1 year follow-up | ||
| Urbanicity of current residence, n (column %) | ||
| Rural | 246 (26.8%) | 1,460 (26.3%) |
| Urban | 412 (44.9%) | 2,376 (42.8%) |
| Suburban | 260 (28.3%) | 1,722 (31.0%) |
| U.S. Census geographic region of residence, n (column %) | ||
| West | 94 (10.3%) | 1,155 (20.8%) |
| Midwest | 152 (16.7%) | 913 (16.5%) |
| Northeast | 325 (35.6%) | 1,368 (24.7%) |
| South | 342 (37.5%) | 2,111 (38.1%) |
Multivariate Logistic Regression Model for Homeboundness at 1-year Post-injury
We identified a minimum set of 19 variables from a list of 39 candidate variables using LASSO in the training subsample to predict year-1 homeboundness in the independent testing subsample. We have provided the order of variables retained and excluded in Supplemental Table 2
In the testing subsample (see Table 2), we found that the following variables were significantly associated with homebound status at 1-year post-injury: older age, having less than a Bachelor’s degree, having Medicaid (versus private) insurance, living in the Northeast or Midwest (versus West), riding with someone else or a special bus as a primary mode of transportation (versus driving oneself), being retired or unemployed (versus being employed or a student), and requiring assistance (versus no assistance) for locomotion, upper body dressing, and social interactions at 1-year post-injury. Other variables (i.e., sex, race, living arrangement, urbanicity, history of prior TBI, pre-injury limitations in getting out of the house, assistance with: problem solving, bladder and bowel) were among the 19 variables selected into the model, but were not statistically significant predictors in the independent testing subsample.
Table 2.
Multivariable Modified Poisson Regression Model with Robust Standard Errors for Year-1 homeboundness post-injury (using the testing dataset; n=2,964, n=329 homebound, n=2,321 not homebound)
| Relative Risk (95% CI) | P-value | |
|---|---|---|
| Age (per 10-year increase) | 1.13 (1.05, 1.20) | 0.001* |
| Sex (male vs. female) | 0.91 (0.75, 1.10) | 0.309 |
| Race/ethnicity | ||
| White | Reference | |
| Black | 1.00 (1.77, 1.30) | 0.989 |
| Hispanic | 0.98 (0.73, 1.32) | 0.910 |
| Other | 1.29 (0.84, 1.97) | 0.240 |
| Education | ||
| Less than Bachelor’s degree | 1.76 (1.29, 2.39) | <0.001* |
| Bachelor’s degree or greater | Reference | |
| Inpatient rehabilitation primary payer source | ||
| Private insurance | Reference | |
| Medicare | 1.11 (0.82, 1.49) | 0.509 |
| Medicaid | 1.43 (1.11, 1.85) | 0.006* |
| No fault auto/worker’s compensation | 1.05 (0.73, 1.50) | 0.790 |
| Other | 1.41 (1.00, 2.02) | 0.055 |
| Living arrangement | ||
| Alone | 1.05 (0.77, 1.43) | 0.772 |
| Spouse or Significant Other | Reference | |
| Parents, siblings, other family | 0.97 (0.78, 1.20) | 0.784 |
| Someone else not specified | 0.93 (0.60, 1.44) | 0.749 |
| History of prior TBI | 0.99 (0.80, 1.23) | 0.923 |
| Urbanicity of personal residence | ||
| Suburban | Reference | |
| Rural | 0.99 (0.78, 1.27) | 0.980 |
| Urban | 0.87 (0.84, 1.97) | 0.246 |
| U.S. Census geographic region of residence | ||
| West | Reference | |
| Midwest | 1.72 (1.18, 2.49) | 0.004* |
| Northeast | 1.96 (1.40, 2.75) | <0.001* |
| South | 1.26 (0.89, 1.78) | 0.195 |
| The primary mode of transportation | ||
| Drives themselves | Reference | |
| Rides with someone | 2.48 (1.78, 3.46) | <0.001* |
| Uses public transit | 1.18 (0.71, 1.96) | 0.514 |
| Uses a special bus | 2.40 (1.45, 3.97) | 0.001* |
| No motor transit | 1.48 (0.54, 4.05) | 0.442 |
| Pre-injury limitations getting out of house | 1.07 (0.77, 1.48) | 0.696 |
| Time to follow motor commands | 1.01 (0.99, 1.03) | 0.365 |
| Current employment | ||
| Employed or a student | Reference | |
| Retired | 3.71 (2.14, 6.44) | <0.001* |
| Unemployed or not otherwise classified | 3.43 (2.01, 5.87) | <0.001* |
| Current assistance required with locomotion, either walking or wheelchair (vs. independence) | 1.47 (1.13, 1.91) | 0.004* |
| Current assistance required with upper body dressing (vs. independence) | 1.33 (1.06, 1.66) | 0.015* |
| Current assistance required for problem solving (vs. independence) | 0.94 (0.74, 1.18) | 0.566 |
| Current assistance required for bladder accidents (vs. independence) | 1.11 (0.85, 1.43) | 0.447 |
| Current assistance required with bowel assistance (vs. independence) | 1.15 (0.91, 1.46) | 0.248 |
| Current assistance required with social interactions (vs. independence) | 1.31 (1.06, 1.62) | 0.013* |
Note: The following variables were considered for inclusion the 10-fold cross validation model selection process but their estimates were shrunk to zero using the Adaptive LASSO Regression procedure (using mutually-exclusive the training dataset): Age2, year of interview, acute care length of stay, inpatient rehabilitation length of stay, mechanism of injury, cranial surgery status, bathing independence, bed transfer independence, no assistance with bladder, no accidents with bowel, dressing lower body independence, expression independence, feeding independence, grooming independence, memory independence, independence with stairs, toilet transfer independence, tub transfer independence, toileting independence, and Charlson Comorbidity Index
indicates statistical significance at α=0.05
Risk for Mortality and Nursing Home Entry Up to 5 Years Post-injury
We first evaluated associations between homeboundness 1-year post-injury and mortality risk. The 5-year mortality rate in the Year-1 homebound group was 22.8% (161/707) versus 8.7% (399/4,614) in the Year-1 non-homebound group. The Kaplan Meier plot is provided in Figure 2, and the log-rank test indicated strong evidence for difference in five-year survival curves by homebound status (p<0.001).
Figure 2. Kaplan Meier survival curve for 5-year survival by Year-1 homebound status (n=5,165).
The five year mortality curves by Year-1 post-injury homeboundness status. The log-rank statistic, which indicates whether the survival curves significant vary, was statistically significant (p=<0.0001). The Year-1 homebound group had a shorter time to death than the non-homebound group.
In Cox proportional hazards models (Table 3), we determined homeboundness was associated with increased risk for 5-year mortality (fully adjusted HR=1.69 (95% CI: 1.35, 2.11). In sensitivity subgroup analyses (Supplemental Table 3), the association between homeboundness and mortality was observed for both participants age >65 (HR=1.61, 95% CI: 1.20, 2.15) and age <65 (HR=1.73, 95% CI: 1.22, 2.46). The association between Year-1 homeboundness and Year-2 or Year-5 nursing home entry was significant after accounting for demographic factors (Model 3), but became marginally non-significant after adjustment for clinical factors in the fully adjusted Model 4 (RR=1.90, 95% CI: 0.94, 3.87, p=0.075).
Table 3.
Association¥ between homeboundness, mortality and nursing home entry after TBI
| Unadjusted | Model 1a | Model 2b | Model 3C | Model 4d | ||
|---|---|---|---|---|---|---|
| Comparison |
HR (95%
CI) |
HR (95%
CI) |
HR (95%
CI) |
HR (95%
CI) |
HR (95%
CI) |
|
| 5-year Mortality Model (n=4,633) | Homebound (n=601) vs. Non-Homebound at Year 1 (n=4,032) | 2.76 (2.23, 3.40) | 1.83 (1.48, 2.27) | 1.90 (1.53, 2.35) | 1.94 (1.56, 2.41) | 1.69 (1.35, 2.11) |
|
RR (95%
CI) |
RR (95%
CI) |
RR (95%
CI) |
RR (95%
CI) |
RR (95%
CI) |
||
| Year 2 or Year 5 post-injury Nursing Home┼ Model (Among survivors) (n=4,015) | Homebound (n=465) vs. Non-Homebound at Year 1 (n=3,550) | 3.73 (2.00, 6.95) | 2.58 (1.40, 4.75) | 2.43 (1.31, 4.51) | 2.33 (1.22, 4.48) | 1.90 (0.94, 3.87) |
Acrynoms: Hazard ratio, HR; Relative risk, RR
Model 1: adjusted for age only
Model 2: adjusted for age, sex
Model 3: adjusted for age, sex, race/ethnicity, education
Model 4: adjusted for age, sex, race/ethnicity, education, FIM Motor at discharge, inpatient rehabilitation length of stay, and Charlson Comorbidity Index
Includes nursing home and sub-acute care at Year 2 and/or Year 5 post-TBI
All models adjusted for inverse probability weights (IPW) for selection bias to do informative censoring. The sample weighted back to was the Aim 1 sample (n=6,595) when homeboundness was collected. The variables in the IPW model also included: age, sex, race/ethnicity, education, FIM Motor at discharge, inpatient rehabilitation length of stay, and Charlson Comorbidity Index
Discussion:
The current study provides key data on homeboundness, a concept scantly documented in the TBI population to date. We found that 14.2% of non-institutionalized adult TBI survivors were homebound 1-year post-injury, highlighting a subgroup in need of additional support after inpatient rehabilitation. We highlight the most important risk factors for becoming homebound 1-year post-injury, and its consequences for future risk of mortality and nursing home entry among those who were previously community-dwelling. Our findings have implications for discharge planning from inpatient rehabilitation and for long-term community integration after moderate-to-severe TBI.
Prior literature on homebound populations focuses on older adults, persons living with dementia, and recipients of home-based primary care.28 In our TBI sample, we observed that homeboundness was associated with older age at injury, but identified risk factors extend beyond age – including, for example, reliance on others for driving and locomotion. Although preparing families for life outside the hospital is a focus of inpatient rehabilitation discharge planning, there is often limited support for families post-discharge.29 Considerable responsibilities fall to caregivers who have been thrust into the caregiving role by the sudden, traumatic nature of the injury,30 including things like non-emergency transportation resources (e.g., paratransit) that are neither universally available nor covered via insurance, and therefore confer considerable out-of-pocket expenses.31 We also found that lower socioeconomic status (i.e., Medicaid beneficiaries) was associated with homeboundness, as reported in other populations.32 Additionally, we observed that individuals who required locomotion assistance (wheelchair or walking) were more likely to be homebound, suggesting that a need for physical assistance may contribute to the cost/burden of leaving the home. To our surprise, medical comorbidity burden was not among the principal factors selected in our training dataset associated with homeboundness. It is possible this is due to inadequate sensitivity of our chosen index of medical comorbidity (i.e., Charlson Comorbidity Score calculated using inpatient hospital diagnosis codes) to burden of diseases associated with homebound status, as the Charlson score was developed specifically to predict mortality risk.33 Without a better alternative in the current study, we cannot rule out the role of comorbid disease burden in risk of homeboundness. Future studies with more granular medical information (i.e., linked Medicare utilization claims) may reveal informative associations.
Our findings indicated that individuals requiring assistance for social interactions were more likely to be homebound, suggesting that risks of homeboundness extend beyond physical or transportation barriers. TBI commonly leads to behavioral and emotional impairments that impact social interactions.34-36 Some TBI survivors prefer to hide their deficits from their friends or family due to stigma or out of concern for being a burden, which may contribute to social avoidance.37 In recent years, social isolation and loneliness after TBI have gained recognition.37-39 These concepts are seldom discussed during discharge planning but may be additional topics warranting attention.
Rehabilitation professionals, persons with TBI, and their caregivers identify community/social participation as a critical component of overall well-being and a primary indicator of successful outcomes following injury.40-42 Though discharge home – rather than an institutional setting – is prioritized by numerous invested groups (i.e., patients, families, providers, payers), it may be short-sighted in some instances to consider home discharge a terminal outcome. Individuals with TBI participate less frequently in social and recreational activities compared to their non-injured peers, and – when they do participate – engage more often in sedentary and solitary activities leading to social isolation and adverse mental and physical health consequences.14,43-46 A recent network analysis46 showed the single PART-O item used herein to define homeboundness was the most central item in the entire network, with dense connections within its own subscale (i.e., Out and About) and also with items on other PART-O subscales (i.e., Social Relations and Productivity). Homeboundness is centrally related to all other aspects of meaningful community participation following TBI.
There has been a shift in focus in the rehabilitation field away from an Independence Model toward an Interdependence Model.47 The latter – an Interdependence Model48 – promotes interventions that maximize social and community participation by empowering persons living with a disability such as TBI to engage meaningfully in their communities (i.e., participation enfranchisement49) to build social capital. Those with more social capital are more likely to be healthier and are at lower risk for mortality.47 The findings of the present study suggest that persons who are homebound after TBI may be a particularly vulnerable group who could benefit from rehabilitation interventions to facilitate community integration.
We found that being homebound 1-year after TBI is a significant predictor for future risk of mortality. Being homebound may itself serve as a proxy variable for a range of demographic and health indicators that are associated with death (e.g., older age, physical health problems). However, in incrementally adjusted regression models for age, functional motor limitations, injury severity, and medical comorbidities, the effect of homeboundness on mortality remained robust. Furthermore, our sensitivity analyses confirmed our mortality findings in both older and younger subgroups with TBI. It is noted that the association between homeboundness and nursing home entry from community living did attenuate after adjustment for clinical factors, which suggest that factors like motor functioning and comorbidity burden may have shared variance with homebound status as it applies to risk of transitioning from community-living to a nursing home. Nonetheless, individuals who are homebound – perhaps due to poorer physical health and social isolation and a lack of engagement in meaningful activities outside of the home – may experience a poorer long-term prognosis, particularly for mortality risk, relative to other community-dwelling peers with TBI who are not homebound.
Furthermore, the findings from our prediction model may be helpful to screen participants at risk of homeboundness at earlier time points after injury. We have identified several potentially modifiable factors (e.g., transportation, employment, locomotion assistance, social functioning) that may be highly specific targets for intervention from rehabilitation professions, including vocational rehabilitation, neuropsychology, physical/occupational therapy. Though these providers have long serviced persons with TBI, their goals may or may not have included prioritizing getting patients with TBI outside the home regularly. Our data showing the consequences of becoming homebound on mortality and nursing home risk indicates this should be a priority.
Limitations
This study has limitations. Data were not available to identify specific reasons why persons were homebound, e.g., personal preference or physical limitations. Further research is needed to understand the reasons for homeboundness after TBI. Our sample required inpatient rehabilitation for moderate-severe TBI, and findings may not generalize to the broader TBI population, or those who were lost to follow-up before year 1 (when homeboundness was assessed). Our post-hoc investigation (Supplemental Table 4) suggests those lost more often belonged to demographically vulnerable groups. Our operational homebound definition was based only on pre-pandemic interviews to avoid confounding by stay-at-home orders. It is unclear whether the COVID-19 pandemic had lasting impacts on homebound risk after TBI, and future research should explore potential enduring effects.
We did not explore mechanisms in the current study that may have mediated the relationship between homeboundness and mortality and nursing home entry. This is an important next step of research. Future studies would benefit from identifying modifiable and time sensitive mediating variables of the relationship observed in the present study between being homebound and risk of mortality and nursing home entry after TBI. Candidate mediators to evaluate from the existing homebound literature may include: psychological factors50 (i.e., depression, loneliness), medical factors18 (i.e., chronic diseases, including obesity), and neighborhood factors51 (i.e., walkability).
Conclusions
The present study is the first to describe the phenomena of homeboundness in TBI populations. Compelling evidence suggests that becoming homebound is common after TBI and is associated with poor long-term prognosis across the age spectrum. Findings from this study suggest that those who are living with TBI and are homebound in the community should be prioritized for education, access to resources, and interventions to optimize community participation and combat social isolation.
Supplementary Material
Funding acknowledgements:
National Institute on Disability, Independent Living, and Rehabilitation Research (NIDILRR) (RGK, NDS, and KDO: 90DPTB0028, 90DPTB0009, 90ARHF0008; SBJ: 90DPTB0025), National Institute of Health’s Eunice Kennedy Shriver National Institute of Child Health and Human Development & the National Center for Medical Rehabilitation Research (NCMRR) (RGK: K99HD106060-02). Health Services and Outcomes Research for Aging Populations Training Program funded by the National Institute on Aging (MLP: T32AG066576). National Institute on Aging (KAO: R01 AG060967)
Abbreviations in the manuscript:
- TBI
Traumatic brain injury
- ED
Emergency Department
- PART-O
Participation Assessment with Recombined Tools-Objective
- ICD
International Classification of Disease
- LASSO
least absolute shrinkage and selection operator
- HR
Hazard ratio
- OR
Odds Ratio
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Presentation of this material: Portions of this paper have been presented at annual meetings for the American Congress of Rehabilitation, Gerontological Society of America, the Brain Injury Association of New York State, and the Brain Injury Alliance of Connecticut.
Conflicts: The authors have no conflicts of interest to report.
References:
- 1.James SL, Theadom A, Ellenbogen RG, et al. Global, regional, and national burden of traumatic brain injury and spinal cord injury, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016. The Lancet Neurology. 2019;18(1):56–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Medicare Home Health Coverage. Accessed July 17th, 2023. https://www.medicare.gov/coverage/home-health-services.
- 3.Chen A, Chan V, Zagorski B, Parsons D, Colantonio A. Factors associated with living setting at discharge from inpatient rehabilitation after acquired brain injury in Ontario, Canada. 2014; [DOI] [PubMed] [Google Scholar]
- 4.Eum RS, Brown AW, Watanabe TK, et al. Risk factors for institutionalization after traumatic brain injury inpatient rehabilitation. Journal of head trauma rehabilitation. 2017;32(3):158–167. [DOI] [PubMed] [Google Scholar]
- 5.Evans E, Gutman R, Resnik L, et al. Successful community discharge among older adults with traumatic brain injury in skilled nursing facilities. The Journal of head trauma rehabilitation. 2021;36(3):E186–E198. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Nalder E, Fleming J, Foster M, Cornwell P, Shields C, Khan A. Identifying factors associated with perceived success in the transition from hospital to home after brain injury. The Journal of head trauma rehabilitation. 2012;27(2):143–153. [DOI] [PubMed] [Google Scholar]
- 7.Turner BJ, Fleming JM, Ownsworth TL, Cornwell PL. The transition from hospital to home for individuals with acquired brain injury: a literature review and research recommendations. Disability and rehabilitation. 2008;30(16):1153–1176. [DOI] [PubMed] [Google Scholar]
- 8.Dams-O'Connor K, Juengst SB, Bogner J, et al. Traumatic brain injury as a chronic disease: insights from the United States traumatic brain injury model systems research program. The Lancet Neurology. 2023;22(6):517–528. [DOI] [PubMed] [Google Scholar]
- 9.Masel BE, DeWitt DS. Traumatic brain injury: a disease process, not an event. Journal of neurotrauma. 2010;27(8):1529–1540. [DOI] [PubMed] [Google Scholar]
- 10.Wilson L, Stewart W, Dams-O'Connor K, et al. The chronic and evolving neurological consequences of traumatic brain injury. The Lancet Neurology. 2017;16(10):813–825. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ornstein KA, Leff B, Covinsky KE, et al. Epidemiology of the homebound population in the United States. JAMA internal medicine. 2015;175(7):1180–1186. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Musich S, Wang SS, Hawkins K, Yeh CS. Homebound older adults: Prevalence, characteristics, health care utilization and quality of care. Geriatric Nursing. 2015;36(6):445–450. [DOI] [PubMed] [Google Scholar]
- 13.Cudjoe TK, Prichett L, Szanton SL, Roberts Lavigne LC, Thorpe RJ Jr. Social isolation, homebound status, and race among older adults: Findings from the National Health and Aging Trends Study (2011–2019). Journal of the American Geriatrics Society. 2022;70(7):2093–2100. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Holt-Lunstad J, Smith TB, Baker M, Harris T, Stephenson D. Loneliness and social isolation as risk factors for mortality: a meta-analytic review. Perspectives on psychological science. 2015;10(2):227–237. [DOI] [PubMed] [Google Scholar]
- 15.Pomeroy ML, Cudjoe TK, Cuellar AE, et al. Association of social isolation with hospitalization and nursing home entry among community-dwelling older adults. JAMA internal medicine. 2023;183(9):955–962. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Wang F, Gao Y, Han Z, et al. A systematic review and meta-analysis of 90 cohort studies of social isolation, loneliness and mortality. Nature human behaviour. 2023;7(8):1307–1319. [DOI] [PubMed] [Google Scholar]
- 17.Sander AM, Clark A, Pappadis MR. What is community integration anyway?: defining meaning following traumatic brain injury. The Journal of head trauma rehabilitation. 2010;25(2):121–127. [DOI] [PubMed] [Google Scholar]
- 18.Qiu WQ, Dean M, Liu T, et al. Physical and mental health of homebound older adults: an overlooked population. Journal of the American Geriatrics Society. 2010;58(12):2423–2428. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Xiang X, Brooks J. Correlates of depressive symptoms among homebound and semi-homebound older adults. Journal of gerontological social work. 2017;60(3):201–214. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Whiteneck GG, Dijkers MP, Heinemann AW, et al. Development of the Participation Assessment with Recombined Tools-Objective for use after traumatic brain injury. Archives of physical medicine and rehabilitation. 2011;92(4):542–551. [DOI] [PubMed] [Google Scholar]
- 21.Great Data: Rural U, Suburban Data: Geographical Assignment by Zip Code. Accessed June 1st, 2023. https://greatdata.com/product/urban-vs-rural
- 22.Shortreed SM, Ertefaie A. Outcome-adaptive lasso: variable selection for causal inference. Biometrics. 2017;73(4):1111–1122. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Walter S, Tiemeier H. Variable selection: current practice in epidemiological studies. European journal of epidemiology. 2009;24:733–736. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Zheng Q, Gallagher C, Kulasekera K. Robust adaptive lasso for variable selection. Communications in Statistics-Theoiy and Methods. 2017;46(9):4642–4659. [Google Scholar]
- 25.Tibshirani R. Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society Series B: Statistical Methodology. 1996;58(1):267–288. [Google Scholar]
- 26.Brooks JC, Strauss DJ, Shavelle RM, Paculdo DR, Hammond FM, Harrison-Felix CL. Long-term disability and survival in traumatic brain injury: results from the National Institute on Disability and Rehabilitation Research Model Systems. Archives of physical medicine and rehabilitation. 2013;94(11):2203–2209. [DOI] [PubMed] [Google Scholar]
- 27.Hernán MA, Hernández-Díaz S, Robins JM. A structural approach to selection bias. Epidemiology. 2004;15(5):615–625. [DOI] [PubMed] [Google Scholar]
- 28.Reckrey JM, Kim PS, Zhao D, et al. Care disruptions among the homebound during the COVID- 19 pandemic: An analysis of the role of dementia. Journal of the American Geriatrics Society. 2022;70(12):3585–3592. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Dams-O’ Connor K, Landau A, Hoffman J, St De Lore J. Patient perspectives on quality and access to healthcare after brain injury. Brain injury. 2018;32(4):431–441. [DOI] [PubMed] [Google Scholar]
- 30.Lieshout K, Oates J, Baker A, et al. Burden and preparedness amongst informal caregivers of adults with moderate to severe traumatic brain injury. International Journal of Environmental Research and Public Health. 2020;17(17):6386. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Medicare Non-emergency medical transportation. Accessed March 5, 2024. https://www.cms.gov/medicare/medicaid-coordination/states/non-emergency-medical-transportation
- 32.Ornstein KA, Garrido MM, Bollens- Lund E, et al. The association between income and incident homebound status among older Medicare beneficiaries. Journal of the American Geriatrics Society. 2020;68(11):2594–2601. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Charlson ME, Carrozzino D, Guidi J, Patierno C. Charlson comorbidity index: a critical review of clinimetric properties. Psychotherapy and psychosomatics. 2022;91(1):8–35. [DOI] [PubMed] [Google Scholar]
- 34.Hanks RA, Temkin N, Machamer J, Dikmen SS. Emotional and behavioral adjustment after traumatic brain injury. Archives of Physical Medicine and Rehabilitation. 1999;80(9):991–997. [DOI] [PubMed] [Google Scholar]
- 35.Mattson AJ, Levin HS. Frontal lobe dysfunction following closed head injury. A review of the literature. The Journal of nervous and mental disease. 1990;178(5):282–291. [DOI] [PubMed] [Google Scholar]
- 36.Wood RL, Thomas RH. Impulsive and episodic disorders of aggressive behaviour following traumatic brain injury. Brain Injury. 2013;27(3):253–261. [DOI] [PubMed] [Google Scholar]
- 37.Lowe N, Crawley L, Wilson C, Waldron B. ‘Lonely in my head’: The experiences of loneliness in individuals with brain injury. British journal of health psychology. 2021;26(2):444–463. [DOI] [PubMed] [Google Scholar]
- 38.Kumar RG, Ornstein KA, Bollens- Lund E, et al. Lifetime history of traumatic brain injury is associated with increased loneliness in adults: A US nationally representative study. International journal of geriatric psychiatiy. 2020;35(5):553–563. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Salas CE, Rojas-Líbano D, Castro O, et al. Social isolation after acquired brain injury: Exploring the relationship between network size, functional support, loneliness and mental health. Neuropsychological rehabilitation. 2022;32(9):2294–2318. [DOI] [PubMed] [Google Scholar]
- 40.Cicerone KD. Participation as an outcome of traumatic brain injury rehabilitation. The Journal of Head Trauma Rehabilitation. 2004;19(6):494–501. [DOI] [PubMed] [Google Scholar]
- 41.Pomeroy JML, Sanchez JO, Cai C, et al. Incorporating the concept of relevance in clinical rehabilitation research and its reviews may improve uptake by stakeholders. American journal of physical medicine & rehabilitation. 2022;101(8):775–781. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Schipper K, Visser-Meily JM, Hendrikx A, Abma TA. Participation of people with acquired brain injury: insiders perspectives. Brain injuiy. 2011;25(9):832–843. [DOI] [PubMed] [Google Scholar]
- 43.Berkman LF, Glass T, Brissette I, Seeman TE. From social integration to health: Durkheim in the new millennium. Social science & medicine. 2000;51(6):843–857. [DOI] [PubMed] [Google Scholar]
- 44.Matthews CE, George SM, Moore SC, et al. Amount of time spent in sedentary behaviors and cause-specific mortality in US adults. The American journal of clinical nutrition. 2012;95(2):437–445. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Taylor HO, Taylor RJ, Nguyen AW, Chatters L. Social isolation, depression, and psychological distress among older adults. Journal of aging and health. 2018;30(2):229–246. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.West SJ, Klyce DW, Perrin PB, et al. A network analysis of the PART-O at 1 and 2 years after TBI: a Veterans Affairs Model Systems study. The Journal of head trauma rehabilitation. 2023;38(5):401–409. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.White GW, Lloyd Simpson J, Gonda C, Ravesloot C, Coble Z. Moving from independence to interdependence: A conceptual model for better understanding community participation of centers for independent living consumers. Journal of Disability Policy Studies. 2010;20(4):233–240. [Google Scholar]
- 48.Condeluci A. Interdependence: The route to community. Routledge; 2019. [Google Scholar]
- 49.Heinemann AW, Magasi S, Bode RK, et al. Measuring enfranchisement: importance of and control over participation by people with disabilities. Archives of Physical Medicine and Rehabilitation. 2013;94(11):2157–2165. [DOI] [PubMed] [Google Scholar]
- 50.Kim Y-R, Jung H-S. Effects of social interaction and depression on homeboundness in community-dwelling older adults living alone. International Journal of Environmental Research and Public Health. 2022;19(6):3608. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Twardzik E, Clarke P, Judd S, Colabianchi N. Neighborhood participation is less likely among older adults with sidewalk problems. Journal of aging and health. 2021;33(1-2):101–113. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.