Would You Rather: Quantifying TBI Survivor Perceptions of Functional Status via Their Surrogates

Introduction

Traumatic brain injury remains a major public health burden, with chronic and/or lifelong consequences for many individuals in the prime of their lives.(1) Long-term recovery after hospitalized TBI is variable and difficult to predict with existing tools, leading to their infrequent use in clinical practice.(1–5) Post-TBI recovery and disability are commonly assessed using the validated TBI clinical trial endpoint, the Extended Glasgow Outcome Scale (GOSE).(6–9) The GOSE is an eight-tiered scale that organizes outcomes based on survival, level of consciousness, work status, and return to independence.(10, 11) However, the GOSE is also nonlinear and nonparametric, rendering it difficult to interpret and communicate to patients and families.

A previous public survey of the general population quantified some GOSE states as worse than death after a hypothetical TBI.(12) However, no health utilities exist for disability after an actual TBI. Decision-making built on a hypothetical TBI may inaccurately predict perceptions of a future health status with TBI, which affect providers’ ability to guide patient and surrogate shared decision-making in the acute setting after TBI.(13) TBI survivors and surrogates may rate their quality of life higher than expected, as is the case with survivors of spinal cord injury.(14–16) To address this uncertainty, we aimed to study the opinions of families of actual post-TBI survivors through their available surrogates.

We hypothesized that after a major hospitalization for closed head injury, TBI surrogates via surrogate respondents, would interpret intervals between GOSE health states as non-linear and unequal, with some post-TBI disability states viewed as worse than death and health utility differences between some GOSE states being more pronounced than differences between other GOSE states. We further hypothesized that differences in TBI surrogate demographics, religiosity, and other factors may mediate their perceptions of post-TBI GOSE disability levels.

Methods

Study population

Prior work has focused on perceptions of TBI disability in the general population considering hypothetical closed head injury scenarios.(12) In contrast, for this study we specifically recruited TBI survivors via their adult surrogates, with the head injury having occurred at least one year prior. Surrogates were defined as caregivers and/or decision-makers who had overseen daily care and medical decisions, regularly received updates from their loved one’s medical team, and/or often helped with bathing, clothing, and cooking for their loved one with TBI.

This cross-sectional study recruited participants nationally over 90 days in 2021 via formal email recruitment, social media advertisement, and ResearchMatch, a national web-based recruitment tool designed via the Clinical and Translational Science Awards Consortium and maintained at Vanderbilt University as an Institutional Review Board (IRB) approved data repository. Amazon gift cards were used as incentives. Before answering any research-related questions, participants provided their informed consent. Through Internet Protocol (IP) address and registration restrictions, participants were only able to complete the survey once. The study was approved by our institutional IRB. Supplemental Figure 1 demonstrates a flow chart of the number of respondents who accessed, were eligible, consented to, and completed the survey. Cookies were not used.

Survey Design and Definitions

We developed the electronic survey using the Qualtrics survey platform (Qualtrics, Provo, UT) following EQUATOR guidelines specific to web-based surveys, the Checklist for Reporting Results of Internet E-Surveys (CHERRIES) (17). The CHERRIES guidelines were used to ensure proper reporting of Methods, Results, and Discussion. The survey instrument used in this study, the GOSE Scenario Survey, was derived from the original GOSE structured interview, as previously described.(12, 18) The GOSE Scenario Survey described seven GOSE health states, outlining the central features for each state that determine a person’s disability and functional status, ranging from GOSE 2 (bedridden and unaware, vegetative state) to GOSE 8 (upper good recovery). GOSE 1, used as a comparator in scenarios, corresponds to death. For an alternative outcome one-year post-TBI different than what the actual TBI patient was experiencing, TBI survivors through their surrogates were consecutively exposed to each GOSE health state. Through serial scenarios, surrogates gave preferences for their loved ones’ post-TBI categorical health states across the spectrum from GOSE 2 to GOSE 8. This determined each GOSE health state’s health utility, a numerical value quantifying the quality of life of a health state.(19) Surrogates were specifically asked to answer the questions on behalf of their loved one (i.e., not on their own views for themselves, in contrast to prior work[20]) who had personally experienced a TBI at least one year prior.

We assessed TBI survivors’ health utilities through the filter of the surrogate participants using the standard gamble approach, which is the most widely used technique for measuring quality of life preferences.(21) The standard gamble approach includes a risk element, simulating the inherent uncertainty in medical decision-making.(19) In each scenario, we asked participants to choose between (1) the certainty of continued life in the given health state (e.g., GOSE 4), or (2) taking a gamble with two possible outcomes: a) perfect health, versus b) immediate death. The gamble probabilities were initially set at 50% immediate death, 50% perfect health and then methodically adjusted based on the participants’ responses until the participants reached indifference between the choices offered. The probability of perfect health at the point of indifference is the health utility of that particular GOSE state. Because we accommodated health states that may be considered worse than death (e.g., GOSE 2, vegetative state), the health utility at the point of indifference could be a negative value. Perfect health corresponded to a quality life or health utility score of 1, and a health utility score of 0 was death. As such, the closer the health utility is to 1, the closer the health state is to being valued as full health. Supplemental Figure 2 includes screenshots of the standard gamble approach with the GOSE Scenario Survey using the standard gamble approach.

Once a participant judged a health scenario, he or she advanced to the next scenario and was unable to return to previous scenarios. GOSE health states were presented in random order to control for any carryover effects, and the random scenario order was included as a model covariate. All participants assessed each health state, GOSE 2 through 8. Because death (GOSE 1, i.e., health utility score of 0) was used as a reference state against which other health states were assessed, GOSE 1 was not included in the scenarios presented for utility determination. Most screens had one question per page, and survey length varied depending on participant responses.

Demographic Assessment and Severity of Prior Exposure to TBI

After participants completed the GOSE Scenario Survey, they answered a series of demographic questions, including age, sex, ethnicity, race, education level, marital status, geographic region, employment status, religious strength of faith as measured by the validated Abbreviated Santa Clara Strength of Religious Faith Questionnaire,(22) current health status as measured by the EQ-5D-3L and the EQ-5D Visual Analogue Scale,(23, 24) and prior exposure to traumatic life events using the validated Brief Trauma Questionnaire (BTQ).(25) Demographic questions were specifically placed after the GOSE Scenario Survey to avoid the risk of stereotype threat.(26)

Next, we collected information on the severity of prior exposure to TBI. Surrogate participants were asked to select their current loved one’s post-TBI GOSE state (range: 1–8), choosing from descriptors of the central features for each state that determine disability and functional status. Using descriptors, participants were also asked whether their loved one underwent the following: craniectomy or craniotomy, placement of an intracranial pressure (ICP) monitor, admission to the intensive care unit ICU), ventilator support, tracheostomy, feeding tube placement (excluding temporary nasoenteric feeding), and hospital discharge to a long-term acute care (LTAC) hospital, skilled nursing facility (SNF), or inpatient rehab (IPR).

Survey Exclusions

As previously established, we excluded participant responses for any of the following reasons: incomplete survey responses, all GOSE health states being valued the same, or at least two logical inconsistencies in quality-of-life valuations of GOSE states 2 through 8.(12) A ‘logical inconsistency’ was defined as a participant valuing a more severe health state higher than a less severe health state, such as valuing GOSE 4 (needing daily partial help) above GOSE 6 (can partially resume work).

Statistical Analysis

We calculated median [interquartile range, IQR] health utilities for each GOSE state, ranging from −1 (worse than death) to 1 (full health), with 0 as reference (death, GOSE 1). We determined the association between risk factors and the GOSE utility scores using a proportional odds logistic regression model with GOSE utility scores as the dependent variable (outcome) and GOSE states, sequence presented to participant, actual GOSE, age, sex, marital status, geographic region, years of education (modeled continuously), EQ-5D-3L index score, religious strength of faith, and number of past traumatic life events as the independent variables.

Religious strength of faith was measured using the Santa Clara scale of religiosity (scale: 5–20, with higher equating to more religiosity). The number of past traumatic events was determined using the Brief Trauma Questionnaire (BTQ), ranging from 0–10. Interactions between GOSE states and all other factors were included in the model. An additional exploratory analysis was done to determine factors predictive of higher utility scores for GOSE 4 specifically, given its bimodal distribution. All statistical analyses were done in R version 4.3.1. No statistical weighting was done for the sample included.

Results

Demographics and TBI Survivor Exposures

Of 515 eligible surrogates, 298 surrogates (58%) consented and completed the scenarios on their TBI survivors’ behalf. Surrogates were median aged 46 years, 22% male, 85% Caucasian, and 54% married (Table 1). Surrogates had a median of 15 years education, with 30% in professional employment and 29% in skilled and/or manual employment. Surrogates were distributed across the United States, with 48% from the South and the remainder from the Northeast, Midwest, and West regions. They spanned a range of religiosity, with a median Santa Clara strength of faith scale of 14 [IQR 10, 18], and overall low exposure to traumatic life events.

Table 1.

Characteristics of Surrogates completing GOSE Scenario Surveys on behalf of TBI Patients

Variable	Median [IQR] or n (%)
Age	46 [35, 60]
Male sex	64 (22%)
Hispanic ethnicity	17 (6%)
Caucasian	255 (85%)
Black	11 (4%)
Asian	10 (3%)
Other	17 (6%)
Education in years	15 [14, 18]
Married	160 (54%)
Divorced or separated	42 (14%)
Widow(er)	13 (4%)
Never been married	79 (27%)
EQ-5D visual analogue scale (0–100)	80 [65, 90]
Santa Clara strength of faith scale (5–20)	14 [10, 18]
Brief Trauma Questionnaire scale (0–10)	2 [0, 3]

TBI, traumatic brain injury

GOSE, Extended Glasgow Outcome Scale, measuring functional outcome after TBI IQR, interquartile range

Table 1 includes demographic details and other characteristics of TBI surrogates completing the GOSE scenario surveys on behalf of their family members with a history of TBI at least one year prior. EQ-5D is a five-dimension scale measuring quality of life, ranging from 0 (worst) to 100 (best). The Santa Clara scale of religiosity measures religious strength of faith, ranging from 5 (low) to 20 (high). The Brief Trauma Questionnaire quantifies the number of past traumatic lifetime events, ranging from 0–10.

Surrogates rated their TBI survivors’ current quality of life at a median 60 [IQR 30.3, 78.0] compared to their own quality of life at a median 80 [IQR 65, 90] on a 0–100 scale. TBI survivors had a current median GOSE 5 [3–7], with 53% having undergone craniectomy or craniotomy, 35% having had a feeding tube, and 24% a tracheostomy (Table 2). Full distribution for the cohort’s current GOSE is shown in Supplemental Table 1.

Table 2.

TBI Survivors’ Past Treatments and Current Health Status

TBI Treatment	n (%)
Craniectomy or craniotomy	159 (53%)
ICP monitor placement	121 (41%)
Admission to ICU	221 (74%)
Ventilator support	143 (48%)
Tracheostomy	71 (24%)
Feeding tube placement	103 (35%)
LTAC or SNF placement	89 (30%)
Inpatient rehabilitation placement	166 (56%)
Current Health Status	Median [IQR]
EQ-5D visual analogue scale	60 [30.3, 78.0]
Actual GOSE status	GOSE 5 [3, 7]

GOSE, Extended Glasgow Outcome Scale, measuring functional outcome after TBI

EQ-5D, five-dimension scale measuring quality of life, range 0–100

ICP, intracranial pressure

ICU, intensive care unit

LTAC, long-term acute care facility

SNF, skilled nursing facility

Table 2 summarizes the TBI survivors’ past and current health status as reported by surrogates of family members who experienced TBI at least a year prior. EQ-5D is a five-dimension scale measuring quality of life, ranging from 0 (worst) to 100 (best).

GOSE Health Utility Ratings

Raw health utility scores demonstrated that TBI survivors via their surrogates rated lower GOSE states (equating to more severe disability) as having lower utility, although the relationship was nonlinear and intervals were unequal between states, with a bimodal distribution for GOSE 4 (Figure 1 and Supplemental Figure 3). GOSE 7 and GOSE 8 were rated similarly well, but there was a precipitous drop-off in utility ratings from GOSE 5 to GOSE 4. GOSE 2, GOSE 3, and GOSE 4 were rated worse than death (0 = GOSE 1) by 89%, 64%, and 38% of respondents, respectively (Table 3).

Figure 1.

Extended Box Plots of Utility Scores of TBI Survivors across GOSE levels.

Table 3.

GOSE Utility Scores of TBI Survivors and Percent Rating Worse Than Death

GOSE level	Median [IQR] Utility Score	Percent rating <0 (worse than death)
GOSE 8, good recovery	0.88 [0.50, 0.991	6.0%
GOSE 7, minor deficits	0.85 [0.50, 0.991	6.4%
GOSE 6, can partly resume work	0.75 [0.30, 0.951	14.1%
GOSE 5, unemployed, independent	0.55 [0.04, 0.901	20.1%
GOSE 4, needing daily partial help	0.45 [−0.01, 0.751	37.6%
GOSE 3, bedridden and aware	−0.01 [−0.19, 0.451	64.1%
GOSE 2, bedridden and unaware	−0.01 [−0.30, −0.011	88.6%

GOSE, Extended Glasgow Outcome Scale, measuring functional outcome after TBI

Table 3 depicts the median health utility scores for each post-TBI GOSE state as determined by TBI survivors via their surrogates, as well as the percentage of TBI surrogates who rated each GOSE state as worse than death.

Predictors of GOSE Health Utility Ratings

Surrogate factors predictive of high mean GOSE utility score rates are shown in Figure 2. As the GOSE state increases, the predicted utility score generally increases, meaning that surrogates were likely to rate GOSE 5 higher than GOSE 4 and GOSE 4 higher than GOSE 3 (p<0.0001 for combined GOSE main effect and interaction effects with other variables). The randomized order in which the GOSE scenarios were presented did not have an obvious effect on the GOSE utility score (p=0.3325 for combined order main effect and interaction effect).

Figure 2.

Effect Plot of Factors associated with GOSE Health Utility scores among TBI Survivors.

There was also evidence that surrogate age (p=0.01), surrogate gender (p<0.01), geographic area (p<0.0001 for combined geographic area main effect and interaction effect), and years of education (p<0.01 for combined years of education main effect and interaction effect) impact TBI survivors’ GOSE utility score ratings. Surrogates with more years of education have higher odds of rating rated GOSE 4 greater than average (p=0.02). Compared to the Northeast region, Midwest, South, and West regions tend to have lower expected health utilities. Our modeling indicated no significant relationships between utility rates of GOSE states and current GOSE state of the individual with TBI (p=0.22), meaning that the surrogates of those with lower GOSE functional states were no more likely to rate low GOSE states as unfavorable than surrogates of individuals at higher GOSE functional states. We also detected no significant impact of surrogate EQ-5D (p=0.64), past traumatic experiences as measured using the BTQ (p=0.82), religiosity as measured using the Santa Clara scale (p=0.16) and marital status (p=0.14).

Respondents who rated no GOSE categories as worse than death (n=21, 7%) were more often unemployed (62% vs. 40% in the overall cohort) and their loved one with TBI had a lower current median GOSE level (3 vs. 5 in the overall cohort). We consider these findings exploratory due to the small numbers.

Discussion

Of 298 surrogates close to the experience of an actual lived TBI, 89% felt that their loved one with TBI would view being in a vegetative state (GOSE 2) as worse than death. Similarly, 64% of TBI surrogates judged TBI survivors who were housebound and needing all-day assistance (GOSE 3) as worse than death, and 38% judged needing daily partial assistance (GOSE 4) as worse than death. Of these surrogates, three-quarters had cared for their TBI survivors after an ICU admission, where half had seen their loved one have a craniectomy or craniotomy, and nearly all TBI survivors in this investigation required at least inpatient rehabilitation.

In a prior investigation of public perceptions about hypothetical post-TBI disability, unequally rated health states were similarly demonstrated, with some states perceived as worse than death.(12) The relationship between different GOSE states was also nonlinear and intervals were unequal between health states. In that prior work, over 75% of participants rated a vegetative state (GOSE 2, absence of awareness and bedridden) and about 50% rated lower severe disability (GOSE 3, housebound needing all-day assistance) as conditions worse than death. In contrast to that work, we observed a precipitous drop-off in utility ratings from GOSE 5 (unemployed, independent) to GOSE 4 (needing daily partial assistance). We did redemonstrate years of education and marital status as important predictors of health utility valuations, but we did not observe any difference by race, age, or religiosity of the TBI surrogate. This discrepancy may be explained by the fact that they were asked to respond on behalf of their loved one rather than on their own behalf, or by the smaller sample size in this study. The variation in health utilities by respondent demographic region echoes the well-documented geographic variation in medical decision making, processes of care, and outcomes between TBI centers.(27–29)

The GOSE scale is a widely used albeit imperfect measure of disability and recovery after TBI. The drop-off in perceived health utility that we observed from GOSE 5 to GOSE 4 is particularly relevant as many clinical trials use a dichotomized GOSE outcome, with favorable functional status corresponding to a GOSE of 5 or higher and unfavorable functional status a GOSE 4 or less, despite a paucity of data demonstrating this as a validated cut-off.(30–32) Of the 42 experts participating in the recent Seattle International severe traumatic Brain Injury Consensus Conference (SIBICC), more than half judged certainty of a persistent vegetative state (GOSE 2) or lower severe disability (GOSE 3) to be grounds for a withdrawal of life-sustaining care, whereas 15% felt that upper severe disability (GOSE 4) justified withdrawal of life-sustaining care.(4)

A better understanding of actual stakeholders’ views of the different long-term outcomes after TBI is essential to shared clinical decision-making in the acute setting. Prognostication and shared decision-making with TBI surrogates in the neurocritical care arena is an area of active debate and research.(13, 33, 34) In addition to contributing to the communication with TBI surrogates in the acute setting, the present work contributes to our understanding of preference-based quality of life after TBI and lays the foundation for more accurate statistical comparisons between the effects of various treatment strategies, as well as cost-effectiveness analyses.

Although limited by selection, recruitment, and survivorship biases, our study represents the first rigorous attempt to assess health utilities of disability states after a real-life hospitalized TBI. Our work has several other limitations. We did not differentiate between respondents who were caregivers only and not surrogates, and vice-versa. We did not collect data on the precise relationship between the surrogate caregiver and the TBI survivor, or on whether the respondent and TBI survivor cohabitated, or on caregiver burden metrics. Ultimately, we sought to minimize respondent burden at the conclusion of the GOSE Scenario Survey by only including a limited number of demographic and TBI exposure questions. Sample size also limits the generalizability of our findings and may have resulted in type 2 errors. Nevertheless, we were able to recruit a diverse sample of TBI surrogates from across the U.S. TBI surrogates answered the questions on behalf of their loved ones, and we did not survey TBI survivors directly and query their self-reported quality of life. Surrogates may have different values than the patients themselves, and we cannot exclude the possibility of dishonorable or discordant surrogates. We did not collect data on whether surrogates were patient-designated or next-of-kin. The reasoning behind not recruiting TBI survivors directly was felt to be a larger response bias challenge, in which only higher functioning TBI survivors would have the capacity to respond, as well as the accrual and response challenges from those TBI survivors who have limited ability to think and interact with their environment and our survey. In the multicenter CENTER-TBI observational study, some patients with severe disability had higher quality of life than expected based on their GOSE alone, albeit most of those patients had mild TBI not requiring ICU hospitalization, which is a different TBI sample than this study.(35)

In this index study of TBI survivors who originally required hospitalization resulting in varying levels of disability, neurologic outcomes with lack of independence, GOSE 2–4, were perceived as worse than death by at least one in three via surrogate response. Acknowledging limitations of selection, response, and survivorship biases, these long-term perceptions may spur future study of post-TBI health utilities among TBI survivors and their caregivers, ultimately informing earlier post-TBI shared decision making and bringing the long-term outcomes used in TBI clinical trials in alignment with patient and family stakeholders, as well as help shaping future value-based research and quality metrics.