Trajectories of Recovery Following Traumatic Brain Injury Among Older Medicare Beneficiaries

Abstract

It is well-known that older adults have poorer recovery following traumatic brain injury (TBI) relative to younger adults with similar injury severity. However, most older adults do recover well from TBI. Identifying those at increased risk of poor recovery could inform appropriate management pathways, facilitate discussions about palliative care or unmet needs, and permit targeted intervention to optimize quality of life or recovery. We sought to explore heterogeneity in recovery from TBI among older adults as measured by home time per month, a patient-centered metric defined as time spent at home and not in a hospital, urgent care, or other facility. Using data obtained from Medicare administrative claims data for years 2010-2018, group-based trajectory modeling was employed to identify unique trajectories of recovery among a sample of United States adults age 65 and older who were hospitalized with TBI. We next determined which patient-level characteristics discriminated poor from favorable recovery using logistic regression. Among 20,350 beneficiaries, four unique trajectories were identified: poor recovery (n = 1929; 9.5%), improving recovery (n = 2,793; 13.7%), good recovery (n = 13,512; 66.4%), and declining recovery (n = 2116; 10.4%). The strongest predictors of membership in the poor relative to the good recovery trajectory group were diagnosis of Alzheimer's disease and related dementias (ADRD; odd ratio [OR] 2.42; 95% confidence interval [CI] 2.16, 2.72) and dual eligibility for Medicaid, a proxy for economic vulnerability (OR 5.13; 95% CI 4.59, 5.74). TBI severity was not associated with recovery trajectories. In conclusion, this study identified four unique trajectories of recovery over one year following TBI among older adults. Two-thirds of older adults hospitalized with TBI returned to the community and stayed there. Recovery of monthly home time was complete for most by 3 months post injury. An important sub-group comprising 10% of patients who did not return home was characterized primarily by eligibility for Medicaid and diagnosis of ADRD. Future studies should seek to further characterize and investigate identified recovery groups to inform management and development of interventions to improve recovery.

Introduction

Traumatic brain injury (TBI) results in over 123,000 hospitalizations and 485,000 emergency department visits annually among older adults in the United States, and rates over the last decade have steadily increased in this age group.^1,2 TBI increases risk for comorbidities including depression, insomnia, and dementia, and results in disability and shorter time to nursing home placement among older adults.^3–8 Yet, despite its large public health impact, relatively little is known about the course of recovery following TBI among older adults or the factors that influence recovery.

It is well known that older adults have poorer recovery following TBI relative to younger adults with similar injury severity.^9-11 However, grouping all older adults together ignores underlying heterogeneity in recovery post-injury. In fact, most older adults recover well from TBI.¹² Thus, identifying those at increased risk of poor recovery would help identity appropriate management pathways, facilitate discussions about palliative care or unmet needs, and permit targeted intervention to optimize quality of life or recovery. Unfortunately, there have been relatively few studies on heterogeneity in recovery following TBI and these have focused primarily on younger populations. For example, Gardener and colleagues modeled functional recovery over 6 months following TBI and identified seven unique recovery trajectories ranging from good recovery to death, yet adults >70 years old were excluded from the study.¹³ Other studies that assessed trajectories of cognitive and physical functioning among adults age 16 and older following inpatient rehabilitation for TBI did not analyze older adults separately.^14,15

Administrative claims data provide large sample sizes and can be useful to study recovery among older adults with TBI as demonstrated by prior studies of incident depression, insomnia, and Alzheimer's disease and related dementias (ADRD) following TBI.^16-18 While no single outcome captures all aspects of recovery following TBI, time spent at home and not in a hospital or other health care facility each month may provide useful insight. Home time is a patient-centered measure associated with higher functional status, lower rates of depression, and increased quality of life that also correlates with patient demographic and comorbid characteristics, making it a potentially useful measure to better understand factors that influence recovery following TBI among older adults.^19–25 The objectives of the present study were to identify unique monthly home time recovery subgroups following TBI among older adults and determine which patient-level characteristics discriminate poor from favorable recovery.

Methods

We used Medicare administrative claims and assessment data obtained from the Center for Medicaid and Medicare Services (CMS) Chronic Conditions Warehouse (CCW) for years 2010-2018 to conduct a retrospective cohort study. For years 2010-2016, we obtained a 5% random sample of claims and a 20% random sample for years 2017-2018. Claims were available for hospitals, skilled nursing facilities, home health care agencies, and both freestanding and hospital-based outpatient facilities. This study was exempted from review by the Institutional Review Board of the University of Maryland School of Medicine.

Study population

Medicare fee-for-service beneficiaries age 65 and older who were community dwelling pre-injury, hospitalized with a primary diagnosis of TBI between 2010-2017, and discharged alive were eligible for the study. We excluded beneficiaries discharged to hospice, long-term acute care hospitals, and psychiatric hospitals. Beneficiaries who were not long stay nursing home residents, defined as residence of more than 100 days in a long-term care facility, were considered community dwelling pre-TBI. TBI was defined using International Classification of Disease (ICD) version 9 (ICD-9) and version 10 (ICD-10) codes on inpatient claims and the date of TBI was assigned as the index date. We used the Centers for Disease Control and Prevention (CDC) case definition for TBI for ICD-9 and added two non-specific TBI codes (S09) to the ICD-10 definition.^1,26 Continuous Medicare Parts A and B coverage for 6 months before the index date and for the entirety of time until death or 12-months post-index, whichever came first, was required.

Home time

We adapted the healthy days at home algorithm published by Burke and colleagues to create our home time measure.²⁴ Consistent with our prior work, home time was computed by subtracting the number of days spent in an inpatient, skilled nursing facility, nursing home, emergency department observation, or outpatient observation setting, and the number of days spent deceased, from the total follow-up available for each of the 12 months (monthly level) post-index date.^27-29 Unlike Burke and colleagues' healthy days at home measure, we did not count home health visit days as days away from home.²⁴ Home time was set to missing for all subsequent months following the month of death.

Covariates

Demographic information was obtained from the annual beneficiary summary files. The CCW contains annual flags and dates of first diagnosis since Medicare enrollment for 27 chronic conditions that are identified using CMS algorithms based on specific diagnosis or procedure codes.³⁰ If the date of first diagnosis for a chronic condition occurred prior to the index date, the beneficiary was considered to have that chronic condition. In addition, we captured the original reason for entitlement code, dichotomized as age or disability (including end-stage renal disease) and dual eligibility for Medicare and Medicaid, an indicator of economic disadvantage.

TBI severity measures such as the Glasgow Coma Scale score, a measure of neurological deficit, and the Abbreviated Injury Scale (AIS) score, a measure of anatomic injury severity associated with increased risk of mortality, are not available in administrative claims data.^31,32 To address this limitation, we included a claims-based estimate of the AIS score. The AIS score for each body region is rated on a scale from 1-6, with higher scores indicating greater injury severity and AIS scores of 6/6 designated as unsurvivable. We used the R program IDCPIC-R to covert ICD-9 and -10 codes from the inpatient TBI claim into AIS codes (head region) to provide a measure of anatomical brain injury.³³ The ICDPUC-R has been used in prior studies to estimate the head AIS score in TBI patients.^27,34,35 We also examined length of hospital stay, whether the hospitalization resulted in an intensive care unit stay, and discharge location (home (including home health) versus facility (including skilled nursing facilities and inpatient rehabilitation) as proxies for injury severity.³⁶ Following selection of the appropriate model, we calculated the total sum and the monthly average of home time over the year following injury.

Statistical analysis

Group-based trajectory modeling with Proc TRAJ in SAS (SAS Institute, Cary, NC) was used to model home time trajectories over the twelve months post-TBI and to classify participants into distinct latent trajectory groups. We employed a truncated Gaussian mixture model to analyze our longitudinal data, considering models ranging from two to five groups. The utilization of a truncated Gaussian distribution was particularly advantageous as it allowed for a more accurate representation of the ceiling and floor effects observed in home time.^28,37 We determined the most appropriate model based on a comprehensive evaluation that included both the averaged posterior probability and the Akaike information criterion (AIC).^38,39 To assign each participant to a specific trajectory group, we estimated the posterior probabilities of membership through maximum likelihood estimation.⁴⁰ Each beneficiary was then allocated to the group with the highest predicted probability of membership. Our selection criteria involved choosing the model with the lowest AIC value, along with posterior probabilities indicating an average value of at least around 0.8 within each trajectory group. This model was identified as the optimal choice for our subsequent analysis.

Following selection of the appropriate model, we described the distribution of baseline (i.e., at the time of TBI) clinical, demographic, and injury-related characteristics by trajectory group and tested differences in distributions using chi-squared Goodness of Fit and analysis of variance (ANOVA). We also described the distribution of death and home time during the year post injury and tested distributional differences using the Kruskal Wallis test.

Next, to identify patient-level characteristics that could differentiate individuals more likely to be in the poor versus favorable recovery trajectories, we used logistic regression. We specifically focused on identifying those individuals most likely to experience poor recovery as this group may benefit from more intensive intervention. For this analysis, we included only those assigned to the poor and good recovery trajectories. Backwards selection was used to select the final model with an exit p < 0.15. Odds ratios (OR) and 95% confidence intervals (CI) are reported. These analyses were performed using SAS Studio Release 3.82.

Results

Among 32,560 Medicare beneficiaries with TBI meeting enrollment criteria, 4740 (14.6%) were deceased at discharge, 5659 (17.5%) were discharged to hospice, long-term acute care, or psychiatric hospitals, and 1811 (5.6%) were less than 65 years of age, leaving 20,350 (62.5%) beneficiaries in our study sample.

We chose a four-group model in accordance with the criteria outlined in the methods. The average posterior probabilities of group membership for this four-group model exceeded 79% and are reported in Table 1. We labeled the trajectory groups based on the shape of the trajectories with the assumption that more home time is better. The four groups comprised poor recovery (n = 1929; 9.5%), improving recovery (n = 2793; 13.7%), good recovery (n = 13,512; 66.4%), and declining recovery (n = 2116; 10.4%; Fig. 1). Individuals in the poor recovery group experienced a flat recovery trajectory that hovered near 0 days of home time over the year post-TBI. In this group, the median sum of home time over the year post-TBI was 0 (interquartile range [IQR] 0, 21) days (Table 2). Those with improving recovery experienced initially low home time that slowly approached 30 days of home time per month by about 6 months post-TBI, with median sum of home time at 220 (IQR 18, 278) days. Individuals assigned to the good recovery trajectory had fewer than 30 days of home time per month on average for the first 2 months post-TBI but made a complete recovery of home time by 3 months, with median sum of home time of 347 (IQR 327, 356) days. Finally, those with declining recovery showed an initial improvement in home time that seemed to peak at 3 months post-TBI but then declined, ultimately remaining at an average of fewer than 20 days of home time per month. In this group, the median sum of home time was 187 (IQR 122.0, 263.5) days.

Table 1.

Average Posterior Probabilities of Trajectory Group Membership for the Four-Group Model

	Group 1	Group 2	Group 3	Group 4
Group 1	0.91	0.08	0	0.01
Group 2	0.08	0.79	0.02	0.04
Group 3	0	0.08	0.95	0.08
Group 4	0.01	0.05	0.03	0.87

FIG. 1.

Average monthly time at home over 1 year following hospitalization for traumatic brain injury among older Medicare beneficiaries 2010-2018 by trajectory group, n = 20,350.

Table 2.

Baseline Characteristics of Medicare Beneficiaries Age 65 Years and Older Hospitalized With a Traumatic Brain Injury (TBI) 2010-2017, by Recovery Trajectories of Home Time Over the Year Following TBI, n = 20,350

	Poor recovery, n = 1929	Improving recovery, n = 2793	Good recovery, n = 13,512	Declining recovery, n = 2116	p value
Age	84.2 (7.9)	83.3 (7.9)	80.7 (8.1)	82.4 (8.2)	<0.001
Sex					0.001
Male	771 (40.0)	1281 (45.9)	5290 (43.8)	919 (43.4)
Female	1158 (60.0)	1512 (54.1)	7592 (56.2)	1197(56.6)
Race					<0.001
White	1665 (86.3)	2455 (87.9)	12,042 (89.1)	1870 (88.4)
Black	135(7.0)	163 (5.8)	573 (4.2)	123 (5.8)
Asian	53 (2.8)	62 (2.2)	308 (2.3)	37 (1.8)
Hispanic	37 (1.9)	50 (1.8)	219 (1.6)	36 (1.7)
Other	39 (2.0)	63 (2.3)	370 (2.7)	50 (2.4)
Head AIS* Score					<0.001
1-2	243 (12.6)	302 (10.8)	2045 (15.1)	375 (17.7)
3	592 (30.7)	850 (30.4)	4318 (32.0)	589 (27.8)
4-5	1074 (55.7)	1626 (58.2)	7028 (52.0)	1135 (53.6)
Length of hospital stay					<0.001
< 2 days	86 (4.46)	54 (1.93)	2373 (17.6)	282 (13.3)
2-3 days	506 (26.2)	591 (21.2)	5397 (39.9)	793 (37.5)
4-5 days	475 (24.6)	704 (25.2)	2904 (21.5)	497 (23.5)
> 5 days	862 (44.7)	1444 (51.7)	2838 (21.0)	544 (25.7)
Intensive care unit stay	681(35.3)	1064 (38.1)	4570 (33.8)	709 (33.5)	<0.001
Discharge destination					<0.001
Home	220 (11.4)	248 (8.9)	8562 (63.4)	1004 (47.5)
Facility	1709 (88.6)	2545 (91.1)	4950 (36.6)	1112 (52.6)
Death during follow-up	802 (41.6)	1197 (42.9)	1424 (10.5)	1064 (50.3)	<0.001
Total days at home over year post-TBI, median (IQR)	0 (0, 21)	220 (18, 278)	347 (327, 356)	187 (122, 263.5)	<0.001
Average monthly days at home over year post-TBI, median (IQR)	0 (0, 2.6)	18.8 (6, 23.2)	28.9 (27.3, 29.7)	21 (16.8, 23.8)	<0.001
ADRD	1093 (56.7)	1172 (42.0)	3657 (27.1)	961 (45.4)	<0.001
Anemia	1541 (79.9)	2204 (78.9)	9293 (68.8)	1806 (85.4)	<0.001
Asthma	312 (16.2)	436 (15.6)	2075 (15.4)	477 (22.5)	<0.001
Atrial fibrillation	641 (33.2)	1038 (37.2)	3712 (27.5)	831 (39.3)	<0.001
Cancer	398 (20.6)	654 (23.4)	2801 (20.7)	537 (25.4)	<0.001
Cataracts	1562(81.0)	2336(83.6)	10,544(78.0)	1718(81.2)	<0.001
Chronic kidney disease	954 (49.5)	1400 (50.1)	5119 (37.9)	1280 (60.5)	<0.001
COPD†	808 (41.9)	1147 (41.1)	4392 (32.5)	1015 (48.0)	<0.001
Depression	1059 (54.9)	1420 (50.8)	5756 (42.6)	1203 (56.9)	<0.001
Diabetes	983 (51.0)	1383 (49.5)	5699 (42.2)	1173 (55.4)	<0.001
Heart failure	1027 (53.2)	1494 (53.5)	5333 (39.5)	1282 (60.6)	<0.001
Hip fracture	261 (13.5)	342 (12.2)	950 (7.0)	234 (11.1)	<0.001
Hyperlipidemia	1650 (85.5)	2433 (87.1)	11,566 (85.6)	1851 (87.5)	0.03
Hypertension	1833 (95.0)	2629 (94.1)	12,077 (89.4)	2026 (95.8)	<0.001
Hypothyroid	673 (34.9)	997 (35.7)	4382 (32.4)	811 (38.3)	<0.001
Ischemic heart disease	1394 (72.3)	2110 (75.6)	8682 (64.3)	1683 (79.5)	<0.001
Osteoporosis	701 (36.3)	928 (33.2)	4088 (30.3)	743 (35.1)	<0.001
Rheumatoid/osteoarthritis	1443 (74.8)	2112 (75.6)	9211 (68.2)	1624 (76.8)	<0.001
Stroke	748 (38.8)	1047 (37.5)	3743 (27.0)	814 (38.5)	<0.001
OREC‡					<0.001
Age	1704 (88.3)	2521 (90.3)	12,180 (90.1)	1807 (85.4)
Disability	225 (11.7)	272 (9.74)	1332 (9.86)	309 (14.6)
Dual	1108 (57.4)	1051 (37.6)	2379 (17.6)	726 (34.3)	<0.001

^* Abbreviated injury scale score; ^†Chronic obstructive pulmonary disease; ^‡Original reason for entitlement code.

ADRD, Alzheimer's disease and related dementias.

Table 2 presents characteristics of the sample by trajectory group membership. For ease of description, we compared all groups to the good trajectory group. Relative to beneficiaries experiencing good recovery whose mean age was 80.7 (SD 8.1) years, beneficiaries in the other groups were older (p < 0.001; Table 2). Those in the good recovery group had a lower burden of comorbidities. For example, prevalence of ADRD was 27.1% among those with good recovery compared with 56.7% among those with poor recovery, 42.0% among those with improving recovery, and 45.4% among those with declining recovery (p < 0.001). Those with good recovery were also less likely to have dual eligibility (17.6% vs. 57.4% for those with poor recovery, 37.6% for those with improving recovery, and 34.3% for those with declining recovery, p < 0.001). Those with good recovery were also less likely to have stayed in the hospital for more than five days (21.0% vs. 44.7% for those with poor recovery, 51.7% for those with improving recovery, and 25.7% for those with declining recovery, p < 0.001).

The final logistic regression model predicting membership in the poor relative to the good trajectory group contained all variables reported in Table 3. Relative to those in the good trajectory group, those in the poor trajectory group were significantly more likely to be eligible for Medicaid (OR 5.13; 95% CI 4.59, 5.74) and be diagnosed with ADRD (OR 2.42; 95% CI 2.16, 2.72). Each additional day in the hospital (OR 1.15; 95% CI 1.14, 1.16) was also associated with membership in the poor relative to the good recovery trajectory.

Table 3.

Odds Ratios and 95% Confidence Intervals for Baseline Characteristics Independently Associated With Membership in the Poor Relative to the Good Recovery Trajectory as Measured by Days at Home Over the Year Following Hospitalization for Traumatic Brain Injury Among Medicare Beneficiaries Age 65 Years and Older 2010-2017, n = 15,441

Age	1.05 (1.04, 1.06)
White vs. Other race	1.49 (1.26, 1.77)
Alzheimer's Disease and Related Dementias	2.42 (2.16, 2.72)
Cataracts	0.86 (0.74, 1.00)
Cancer	0.87 (0.76, 1.00)
Depression	1.16 (1.04, 1.30)
Diabetes	1.23 (1.10, 1.38)
Hip fracture	1.35 (1.14, 1.60)
Hyperlipidemia	0.77 (0.65, 0.91)
Hypertension	1.43 (1.11, 1.85)
Stroke	1.20 (1.07, 1.35)
Dual eligibility for Medicaid	5.13 (4.59, 5.74)
Length of stay, per day	1.15 (1.14, 1.16)

Discussion

We identified heterogeneity in recovery of home time among older Medicare beneficiaries following hospitalization for TBI. Four unique trajectories emerged from our analysis, facilitating identification of groups likely to recover well and those which may benefit from targeted intervention. While prior studies have documented poorer recovery among older versus younger adults following TBI, our study is the first to identify unique patterns of recovery of home time following TBI among older adults.^9–12 Home time is not only important to patients, it correlates well with higher functional status, lower rates of depression, increased quality of life, and fewer disabilities in activities of daily living, making it an excellent proxy for recovery across multiple domains of functioning.^19–25,41

One in 10 beneficiaries did not return home over the year post-TBI. Contrary to expectations, Medicaid eligibility and ADRD, rather than TBI severity, were significantly associated with membership in the poor recovery trajectory.⁴² Findings are consistent with prior work showing that economic vulnerability, disability, and cognitive impairment were associated with unsuccessful discharge back to the community following rehabilitation for TBI among older adults.⁴³

While the literature on recovery among older adults with TBI is sparse, hip fracture is a fall-related injury among older adults that has garnered more research attention. In this population, cognitive impairment has been shown to associate with poorer outcomes.^44,45 In a similar trajectory modeling study conducted among patients with hip fracture, Medicaid eligibility was identified as an important factor predicting membership in the poorest home time trajectory.²⁸

The relationship between poverty and dementia is complex because while economic vulnerability increases risk for dementia and disability, economically disadvantaged individuals with ADRD may also be eligible to receive nursing home care through Medicaid, which is a primary payer for nursing home admission after the termination of restorative care interventions.⁴⁶ However, all individuals in this study were community dwelling before TBI, suggesting that entry into long-term care may be due to inadequate supports to promote aging in place for some survivors and a potential target for future interventions. While most older adults prefer living at home to entering long-term care, those with more wealth may be able to choose more costly options to remain in the community such as assisted living while lower income olde adults may only have nursing homes as an option to receive the care they need.

Understanding of recovery and drivers of loss of home time after TBI among older adults is incomplete, but a substantial proportion of mixed-age TBI survivors report unmet needs in activities of daily living or assistance with shopping.^47-49 These needs are likely increased among older adults with TBI. Further, unmet needs are more common for those who have socioeconomic vulnerabilities and higher medical complexity, suggesting a plausible target for future interventions to increase home time for older TBI survivors.

Two-thirds of older adults in this study recovered well following hospitalization for TBI, despite evidence that older adults are less likely to be appropriately triaged, receive less intensive acute treatment and are less likely to receive rehabilitation for TBI compared with younger adults with similar injury severity.^2,42,50–54 This important and hopeful information should bolster efforts to develop guidance for acute management and rehabilitation that is specific for geriatric TBI. Further, given that the majority of TBI among older adults is fall related, fall-prevention among older adults should be strongly emphasized.¹² Although those with consistently good recovery were slightly younger (average age 81 years) than those in the other recovery groups, individuals with improving recovery were among the oldest, with average age of 83 years, suggesting that age is not the only factor driving recovery. While all participants in this study were medically complex, those with higher AIS scores (i.e., more severe TBI) were most likely to spend 5 or more days in the hospital and be discharged to a facility. Nonetheless, when predicting days spent at home over the year following TBI, geriatric factors (i.e., ADRD and comorbid burden) and economic vulnerability may better delineate successful recovery.

While not the primary focus of this study, other trajectory groups should be more thoroughly investigated in future work. Those with declining recovery represented 10% of the population. This group had the highest burden of comorbidities and was most likely to have qualified for Medicare due to disability, a known risk factor for TBI.⁵⁵ They were less likely to be discharged to a facility relative to the poor and improving recovery groups but were the most likely of all groups to die during follow-up. Of all groups, they were also the most likely to have an AIS head score of 1-2, suggesting mild TBI. Additional research is needed to better understand recovery from TBI in this group.

The group with improving recovery represents a success story for recovery post-TBI as measured by home time. This group, 14% of the total sample, was most likely to have severe TBI (e.g., head AIS scores of 4-5, long length of hospital stay, intensive care unit stay, and discharge to facility). Yet, by approximately 6 months post injury, they were spending the majority of time at home. This recovery time frame is consistent with prior studies conducted among newly disabled older adults and hip fracture patients showing that most functional recovery is complete by 6 months post-injury.^56,57

This study has limitations that should be noted. Relative to the larger proportion of older adults diagnosed with TBI in the emergency department, participants in this study may have had more severe injury given hospitalization with TBI was a study criterion.¹ Nonetheless, the majority (66%) recovered well, and another significant proportion (14%) recovered sufficiently to spend most of their time at home by the end of the year. To extend results to milder TBI, future studies could assess home time among older adults diagnosed with TBI in other settings such as the emergency department. Additionally, there are limitations associated with the use of administrative claims data, including lack of clinical severity information and underdiagnosis of certain conditions, particularly ADRD.^58,59 This study was conducted among Medicare fee-for-service beneficiaries. Relative to fee-for-service beneficiaries, those with Medicare Advantage tend to be older, more racially diverse, and of lower socioeconomic status.⁶⁰ These differences could impact the distribution of beneficiaries in the identified recovery trajectories. Future work should investigate recovery following TBI this population. Finally, unique aspects of the healthcare and insurance system in the U.S., particularly administration of Medicare and Medicaid, mean findings may not generalize to other countries/healthcare systems. Investigating recovery following TBI among older adults in non-U.S. locations would be a significant contribution to the literature.

In conclusion, this study identified four unique trajectories of recovery over 1 year following TBI among older adults. Using home time, a patient-centered metric, we found that two-thirds of older adults hospitalized with TBI returned to the community and stayed there. Recovery of home time was complete for most by 3 months post-injury. An important sub-group comprising 10% of patients who did not return home; membership in this group was strongly associated with Medicaid eligibility and diagnosis of ADRD. Future studies should seek to further characterize and investigate identified recovery groups to inform management and development of interventions to improve recovery.

Transparency, Rigor, and Reproducibility Summary

Medicare administrative claims data obtained from the Centers for Medicare and Medicaid Services were used for analysis. These data are available to investigators with an approved data use agreement for a fee. The data use agreement prohibits data sharing. We used the CDC's validated TBI surveillance definition which is diagnosis-code based and commonly used in claims-based studies.¹ We adapted the healthy days at home algorithm published by Burke and colleagues to create our claims-based home time measure, which was computed by subtracting the number of days spent in an inpatient, skilled nursing facility, nursing home, emergency department observation, or outpatient observation setting, and the number of days spent deceased, from the total follow-up available for each of the 12 months (monthly level) post-index date.^24,27-29 Our statistical analysis was conducted using readily available programs (e.g., proc traj, proc logistic) in SAS (SAS Institute, Cary NC) which is available to investigators for a fee. The study and analysis plan were not preregistered. We did not undertake sample size or power calculations due to the very large size of the dataset we obtained. The study was adequately powered. Results from this study should be replicated in other populations.

Authors' Contributions

Jennifer Albrecht: conceptualization, formal analysis, writing—original draft, review, and editing; Chixiang Chen: formal analysis, writing—review and editing; Jason Falvey: Resources, funding acquisition, conceptualization, writing—review and editing.

Funding Information

Dr. Albrecht was supported by National Institute on Aging grant R01AG076441. Dr. Falvey was supported during the work by the National Institute on Aging and the Maryland Claude D. Pepper Center (grant numbers K76AG074926 and P30AG028747). We acknowledge the support of the University of Maryland, Baltimore, Institute for Clinical & Translational Research and the National Center for Advancing Translational Sciences Clinical Translational Science Award grant number 1UL1TR003098 in helping secure Medicare data and providing support for analysis.

Author Disclosure Statement

No competing financial interests exist.