Measuring emotional suppression in caregivers of adults with traumatic brain injury

Abstract

Objective.

Caregivers of individuals with traumatic brain injury (TBI) often feel pressure to maintain the appearance that they are emotionally well-adjusted, despite feelings to the contrary. As there are currently no measures examining this construct, this paper focuses on the development of a new measure that is specific to caregivers of people with TBI.

Design.

533 caregivers of civilians with TBI (n = 218) or service members/veterans (SMVs) with TBI (n = 315) completed 43 emotional suppression items, as well as other patient-reported outcomes and an estimate of the functional ability of the person with TBI.

Results.

Exploratory and confirmatory factor analyses supported the retention of 25 items. Graded response model (GRM) analyses and differential item functioning (DIF) studies supported the retention of 21 items in the final measure. Expert review and GRM calibration data were used to develop a 6-item static short form (SF) and program a computer adaptive test (CAT). Internal consistency was excellent for both the CAT and SF (reliabilities ≥ 0.91); three-week test-retest stability was good (all intraclass correlations ≥ 0.89). Convergent validity was supported by moderate associations between TBI-CareQOL™ Emotional Suppression and related measures (rs from 0.47 to 0.59); discriminant validity was supported by small correlations between Emotional Suppression and positive aspects of caregiving and physical health (rs from 0.14 to 0.28). Known-groups validity was also supported.

Conclusions.

The new TBI-CareQOL Emotional Suppression CAT and 6-item short form is the first self-report measure of this construct in this population. Our findings suggest this new measure has strong psychometric properties.

Caregivers of persons with moderate to severe traumatic brain injury (TBI) experience substantial burden associated with their fulfilling the caregiving role, and this burden often persists over time (Bayen et al., 2016; Manskow et al., 2017). Such burden is experienced by caregivers of Service Members and Veterans (SMVs) with TBI, as well as civilians with TBI (Brickell, French, Lippa, & Lange, 2018; Griffin et al., 2017). These perceptions of burden are reflected in high levels of emotional distress among caregivers of both civilians (Kreutzer et al., 2009; Schonberger, Ponsford, Olver, & Ponsford, 2010; Winstanley, Simpson, Tate, & Myles, 2006) and SMVs (Brickell et al., 2018; Griffin et al., 2017). Research has documented substantial depression and/or anxiety in caregivers as early as three months after beginning care for a person with TBI and persisting up to seven years (Brooks, Campsie, Symington, Beattie, & McKinlay, 1987; Kreutzer et al., 2009; Livingston, Brooks, & Bond, 1985; Ponsford & Schonberger, 2010).

While many caregivers of persons with TBI experience perceived burden and distress, there is variation in perceptions of the caregiving role, as well as in emotional responses to the role. One factor that can influence perceptions of burden and emotional response to the caregiving role is type of coping strategies used. Prior research has shown that emotion-focused coping, or coping that is focused on denying, avoiding, or inhibiting thoughts and emotions, and/or their associated triggers, is associated with greater perceived burden (Hanks, Rapport, & Vangel, 2007) and emotional distress (Davis et al., 2009; Sander, High, Hannay, & Sherer, 1997). In contrast, use of strategies that focus on problem-solving (planning and goal setting to directly target specific problems; Lazarus, 1991) or on accepting emotions rather than attempting to alter, avoid or control them, (Hayes et al., 2006) are associated with better psychological functioning in caregivers.

Emotional suppression can be viewed as an emotion-based coping strategy, as it is focused on inhibiting expression of emotion. Studies have demonstrated that emotional suppression is associated with negative physical effects (e.g., increased blood pressure; Gross & Levenson, 1993; Roberts, Levenson, & Gross, 2008), reduced positive affect (Abler et al., 2010; Gross & John, 2003; Kwon & Kim, 2018; Larsen et al., 2012), and lower life satisfaction (Gross & John, 2003; Kwon & Kim, 2018; Sheldon, Ryan, Rawsthome, & Ilardi, 1997). Emotional suppression is also associated with reduction in one’s ability to attend to and process social information (Egloff, Schmukle, Burns, & Schwerdtfeger, 2006; Hayes et al., 2010; Moore & Zoellner, 2012; Richards & Gross, 2000). The discrepancy between what the person is feeling and his or her presentation to others can contribute to negative feelings, such as poor self-esteem (John & Gross, 2004). In contrast, more positive emotional regulation strategies, such as reappraisal, are generally associated with greater positive emotion and well-being (Gross & John, 2003; John & Gross, 2004; Troy, Saquib, Thal, & Ciuk, 2018).

To date, there has been minimal investigation examining emotional suppression in caregivers of persons with TBI. In a recent qualitative analysis of focus groups conducted with caregivers of SMVs with TBI, caregivers commonly reported suppressing their emotions (Carlozzi et al., 2016). These caregivers felt that they should present a brave face to others, regardless of what they were feeling (“How do you feel?” That’s a question I get a lot. “How do you feel?” I try not to.”). The frequency with which the topic of emotional suppression was raised during these focus groups in SMVs, and to a lesser extent in civilians (Carlozzi et al., 2015a), and its potential for negative effects, highlights the need for a valid measure of caregivers’ suppression of emotions. Unfortunately, there is no existing measure of emotional suppression for clinical populations, including caregivers to advance this line of research. Furthermore, a brief, psychometrically sound measure of emotional suppression could help to guide education and treatment for caregivers and to assess the effectiveness of these initiatives for caregivers of persons with TBI. As such, this study proposes to address this gap through the development of an Emotional Suppression item bank, short form, and computer adaptive test for caregivers of persons with TBI (including both civilians and SMVs). Convergent and discriminant validity were assessed by investigating the relationship between emotional suppression and measures of functional independence in the person with injury, caregivers’ perceived burden, and caregiver health-related quality of life (HRQOL). Caregivers’ emotional suppression was hypothesized to be negatively associated with functional independence in the person with TBI, based on research showing that greater independence in persons with TBI is associated with better caregiver emotional functioning, (Sander, Maestas, Sherer, Malec, & Nakase-Richardson, 2012) and emotional suppression is likely to be associated with less positive affect (Abler et al., 2010; Gross & John, 2003; Kwon & Kim, 2018; Larsen et al., 2012). Emotional suppression was hypothesized to be positively related to perceived burden based on research showing a positive relationship between perceived burden and emotional distress (Struchen et al., 2012). Finally, emotional suppression was hypothesized to be negatively related to caregivers’ HRQOL based on prior findings that emotional suppression is associated with negative physical health and lower life satisfaction (Gross & John, 2003; Kwon & Kim, 2018; Sheldon et al., 1997).

Methods

Study Participants

A total of 533 caregivers participated in this study (218 caregivers of civilians with TBI and 315 caregivers of SMVs with TBI). Hospital and community-based recruitment, existing research registries, and medical record data capture systems (Hanauer, Mei, Law, Khanna, & Zheng, 2015) were used to enroll caregivers of civilians into this multi-site study (i.e., University of Michigan, Baylor College of Medicine/TIRR Memorial Hermann, and the Rehabilitation Institute of Michigan). Hospital and community-based recruitment and site-specific research registries were used to enroll caregivers of SMVs (i.e., James A. Haley Veterans Hospital in Tampa, Walter Reed National Military Medical Center, and the University of Michigan).

Inclusion criteria for both caregiver groups were as follows: Caring for an individual who was ≥ 16 years of age at the time of injury and was currently at least one-year post injury, the ability for the caregiver to read and understand English, an indication that the caregiver provided physical assistance, financial assistance, and/or emotional support to an individual with TBI. Professional (i.e., paid) caregivers were not eligible for this study (Note. This does not include family caregivers that receive financial compensation for their role). Inclusion criteria for caregivers of civilians also included the following: Caring for someone with a medically documented complicated mild, moderate, or severe TBI (based on TBI Model Systems criteria). An additional inclusion criterion for caregivers of SMVs was as follows: Caring for an individual with a medically documented diagnosis of a TBI (documentation must have been from a U.S. Department of Defense or U.S. Department of Veteran Affairs [DoD/VA] treatment facility). When available, medical record documentation was used to classify SMV TBI severity as mild, moderate, severe, or penetrating according to the DoD/VA (The Management of Concussion/mTBI Working Group, 2009). All study activities were conducted in accordance with the local institutional review boards; caregivers provided informed consent prior to participation in this study (or a waiver was granted due to the low-risk nature of this study).

Measures

TBI-CareQOL™ Emotional Suppression Item Pool.

The initial Emotional Suppression item pool was developed to reflect focus group discussion findings from among caregivers of individuals with TBI (Carlozzi et al., 2016; Carlozzi et al., 2015b). An iterative process was used to refine this item pool and included feedback from expert review, item reading level assessment, cognitive interviews, and a final consensus meeting by study team investigators. Figure 1 summarizes this process for the Emotional Suppression item pool. The final item bank yields scores on a T-score metric (M = 50; SD = 10), with higher scores indicating more emotional suppression. For reliability and validity analyses, we examined the full Emotional Suppression item bank as well as T-scores from both the 6-item short form (SF) and computer adaptive test (CAT) simulations, which were conducted using Firestar Version 1.3.2 (Choi, 2009).

Figure 1.

Iterative Process for Emotional Suppression Item Pool

Health-Related Quality of Life.

The Patient Reported Outcomes Measurement Information (PROMIS) system was developed to measure health-related quality of life (HRQOL) in the general population (Cella et al., 2010). For the current study, the PROMIS Social Isolation CAT was administered to examine one’s sense of companionship, communication with others, and sense of belonging. PROMIS Social Isolation has been found to have high internal consistency (α = 0.92 and 0.94 for civilian and military samples, respectively) in caregivers of persons with TBI (Carlozzi et al., 2019). Social Isolation scores are on a T-score metric (M = 50; SD = 10), with higher scores indicating more feelings of isolation. This measure was administered to examine the convergent validity of the final Emotional Suppression item bank.

Caregiver Appraisal.

The Caregiver Appraisal Scale (CAS) assesses positive and negative aspects of the caregiving role (Lawton, Kleban, Moss, Rovine, & Glicksman, 1989). For the purposes of this study, we administered the 35-item version of this measure. Four separate subdomain scores (perceived burden, caregiver relationship satisfaction, caregiving ideology, and caregiving mastery; Struchen, Atchison, Roebuck, Caroselli, & Sander, 2002) were used to examine convergent and discriminant validity of the Emotional Suppression item bank. Internal consistency for the CAS subscales was good for Burden (α = 0.91 and 0.90 for civilians and SMVs, respectively) and Satisfaction (α = 0.78 and 0.82 for civilians and SMVs, respectively), but low for Ideology (α = 0.68 and 0.64 for civilians and SMVs, respectively) and Mastery (α = 0.64 and 0.57 for civilians and SMVs, respectively). Note that the internal consistency for the CAS subscales were similar to those found by Struchen and colleagues (2002).

Physical and Mental Health.

The RAND-12 is a 12-item measure of physical and mental HRQOL. We examined the RAND-12s Physical Health Composite (PHC) and Mental Health Composite (MHC) scores; these scores are on a T-score metric (M = 50; SD = 10), with higher scores indicating better health. Mosier’s (1943) alpha for weighted composite scores was calculated for the current sample, and was good for both PHC (α = 0.90) and MHC (α = 0.74). This measure was administered to examine the convergent validity of the final Emotional Suppression item bank.

Disability of Persons with TBI.

The Mayo Portland Adaptability Inventory 4^th edition (MPAI-4) was developed as a proxy assessment for disability in persons with TBI (to be completed by either a clinician, caregiver, or person with injury; Malec, 2005). This measure is on a T-score metric (M = 50; SD = 10), with higher scores indicating more impairment. Scores on the MPAI-4 were used to divide caregivers into two separate groups, a “high function” group of caregivers caring for persons with T-scores < 60 and a “low function” group for caregivers of persons with T-scores ≥ 60 (Malec, 2005).

Data Collection

All self-report data were collected using assessmentcenter.net. Participants completed assessments on either a personal or publically-available computer with internet access or using a study-specific research computer (for participants without an internet connection). A subset of participants (n = 132) seen at three of the data collection sites (i.e., University of Michigan, TIRR Memorial Hermann, and the Rehabilitation Institute of Michigan) completed test-retest assessments within three weeks of their initial survey.

Statistical Analyses

Item Bank Development.

Item bank development was conducted according to published measurement development standards (Helkala et al., 1995). This process included both classical and item response theory (IRT) analytical approaches. First, exploratory factor analyses (EFA) and confirmatory factor analyses (CFA) and clinical input (Cook, Kallen, & Amtmann, 2009; McDonald, 1999; Reise, Morizot, & Hays, 2007) were used to identify a unidimensional set of items. With regard to EFA, essential unidimensionality was supported if: 1) the ratio of eigenvalue 1 to eigenvalue 2 > 4; and 2) the proportion of variance accounted for by eigenvalue 1 >.40. Items with sparse cells (i.e., response categories with n < 10 respondents) or which had low item-adjusted total score correlations (i.e., < 0.40) were excluded from the item pool. Additionally, non-parametric IRT models examining item-rest plots and expected score by latent trait plots were used to examine item monotonicity (Testgraf Software; Ramsay, Aug 1, 2000). Items found to be non-monotonic were also excluded from the item pool. With regard to CFA, essential unidimensionality was supported by the following model fit criteria: comparative fit index (CFI) ≥ 0.90, Tucker-Lewis index (TLI) ≥ 0.90, and standardized root mean squared residual (SRMR) ≤ 0.08. While literature traditionally suggests that root mean square error of approximation (RMSEA) be <0.08 (excellent fit) or <0.10 (acceptable fit), conflicting studies have found that these cutoffs are too conservative for assessing essential unidimenionality, as number of items, latent score distribution, model specification, degrees of freedom, and sample size directly impact standard CFA fit criteria (Cook et al., 2009; Chen, Curran, Bollen, Kirby, & Paxton, 2008). As such, criterion for the RMSEA was set to < 0.15 for the purposes of the current analyses (Bentler, 1990; Chen, Curran, Bollen, Kirby, & Paxton, 2008; Hatcher, 1994; Hu & Bentler, 1999; Kline, 2005). Items with low factor loadings (Ix < 0.50) and items demonstrating local dependence (residual correlation > 0.20; correlated error modification index ≥ 100; Cook et al., 2009; McDonald, 1999; Reise et al., 2007; Whittaker, 2012) were also removed from the pool. For all identified “locally dependent” item pairs, the item whose content was deemed by study content experts to be most relevant for assessing Emotional Suppression was retained. EFA and CFA were conducted using Mplus version 7.4 (Muthén & Muthén, 2011). EFA modeling used a geomin (oblique) rotation, with an epsilon setting of 0.001 optimized for up to three factors extracted. Both EFA and CFA modeling treated indicators as categorical in nature; both also employed the weighted least squares mean and variance adjusted (WLSMV) parameter estimation method.

We further investigated potential measure multi-dimensionality by conducting bifactor analyses, the results of which were designed to highlight any multi-dimensionality identified and, if warranted, provide support for considering the ES measure as essentially unidimensional. We compared general factor item slope estimates from a unidimensional vs. multi-dimensional (i.e., one general + two specific factors) model to examine important impact from potential multi-dimensionality (i.e., slope absolute value differences ≥ 0.50). We also calculated and examined general factor omega vs. omega-H values for any large discrepancy, indicative of multi-dimensionality impact. In addition, we examined the omega-H values from specific factor models factor 1 and 2 and compared them to the corresponding general factor omega values to determine if omega-H values indicated any substantial reliable variance was associated with a specific factor, suggestive of important multi-dimensionality.

Next, Samejima’s graded response model (GRM; Samejima, van der Liden, & Hambleton, 1996) was used to assess model and item fit and to estimate item parameters. Items were removed if they displayed significant misfit (S-χ², p < 0.01). A final model IRT-based RMSEA overall fit value was also estimated. GRM analyses were conducted in IRTPRO version 3.0 (Cai, Thissen, & du Toit, 2015). Following these analyses, differential item functioning (DIF) studies were conducted to examine potential item bias. DIF was investigated for the following factors: age (≤ 40 vs. > 40 years), education (high school graduate or less vs. > greater than high school), and caregiver group (civilian vs. SMV). Items were excluded if they exhibited impactful DIF, which was defined as follows: (1) items flagged by Nagelkerke pseudo-R² change values ≥ 0.20, and (2) > 2% of DIF-corrected vs. uncorrected score differences exceeded uncorrected score standard errors. DIF was examined using a hybrid IRT ability score-ordinal logistic regression framework (Crane, Gibbons, Jolley, & van Belle, 2006), followed by score impact analyses; item flagging analyses were conducted using the R package LORDIF version 0.3–3 (Choi, Gibbons, & Crane, 2011; R Core Team, 2014). Following these analyses, CFA was again employed to ensure essential unidimensionality of the proposed final item set (using the same fit criteria outlined above). Item calibration estimates from the GRM analyses were used to program the item bank as a computer adaptive test. A 6-item short form was also constructed. These SF items were selected by a consensus process among TBI, caregiver, and measurement development experts. Items were purposely selected to represent the full range of clinical concept coverage, while being psychometrically supported for SF inclusion by their item calibration and calibration-related statistics (e.g., item slope, thresholds, average item difficulty, and, particularly, score-level item information).

Reliability and Validity Analyses.

Assessment of data skewness and kurtosis in both administration formats and caregiver groups met our criteria for normal distribution (skewness and kurtosis < |1|) and was thus appropriate for parametric analyses. These analyses were conducted for the full sample as well as separately for the two different caregiver groups.

For the full Emotional Suppression item bank, IRT-based marginal (internal consistency) reliability was estimated. Cronbach’s alpha (internal consistency) reliability was calculated for the Emotional Suppression SF item responses, and a comparable IRT-based internal consistency (i.e., average standard error-based) reliability was calculated for the Emotional Suppression simulated CAT scores to complete an examination of internal consistency reliability. Two-way mixed intraclass correlations assessing consistency agreement were used to examine test-retest reliability, and Pearson correlations were computed to examine the relationships between the different Emotional Suppression measure administration formats (i.e., full item bank, simulated CAT, and SF). Minimal acceptable reliability was specified as ≥ 0.70 (Cohen, 1988; DeVellis, 2017). We also report the percentage of participants who had the highest possible SF or CAT score (ceiling effect) and the percentage of participants who had the lowest possible SF or CAT score (floor effect). For ceiling and floor effect analyses with simulated CAT scores, we divided the raw summed CAT item responses by the number of items administered (e.g., a score of 5 was a “ceiling effect” and a score of “1” was a “floor effect”). Acceptable floor and ceiling rates were set at ≤ 20% (Andresen, 2000; Cramer & Howitt, 2004). In addition, we report administration times for both CAT and SF versions (based on start and stop times for each item, which were recorded electronically).

Convergent and discriminant validity of the Emotional Suppression item bank were examined using Pearson correlations. Strong correlations (r > 0.6) were expected between Emotional Suppression and Social Isolation (which was hypothesized as the closest comparator among the different HRQOL measures), which were interpreted as being good evidence for convergent validity. Moderate correlations (rs between 0.4 and 0.6) were expected between Emotional Suppression and negative aspects of caregiving (i.e., CAS caregiver burden) and between Emotional Suppression and mental HRQOL (Rand-12 MHC), which would also support convergent validity. Weak correlations (r < 0.3) were expected between Emotional Suppression and physical HRQOL (i.e., Rand-12 PHC) and between Emotional Suppression and positive aspects of caregiving (i.e., CAS caregiver relationship satisfaction, caregiving ideology, and caregiving mastery), which were interpreted as evidence for discriminant validity (Campbell & Fiske, 1959).

Known-groups validity was also examined for Emotional Suppression. For these analyses, independent sample t tests were used to compare caregivers of high vs. low functioning persons with TBI in both civilian and SMV subgroups. Known-groups validity would be supported if caregiver group differences were observed between those caring for low vs. high functioning individuals (i.e., caregivers of persons that were low functioning should report more emotional suppression than caregivers of persons that were high functioning). As a secondary hypothesis, known-groups validity would also be supported if caregiver differences were observed between civilian vs. SMV subgroups (i.e., we expect caregivers of SMVs to report more emotional suppression than caregivers of civilians, due to military culture influence).

Finally, we examined clinical impairment rates (i.e., participants whose scores were > 1 SD worse than the sample mean of 50) to determine if caregivers of persons who were lower functioning were at greater risk for emotional suppression than those who were caring for higher functioning individuals. A Z-test for two proportions was used to identify significant differences in impairment rates across the two groups. In addition, impairment rates that exceed 16% would also indicate high risk (Heaton, Miller, Taylor, & Grant, 2004).

Sample Size Requirements.

Sample size for this study was based on recommendations for the GRM-based analyses and the DIF analyses employed during the item bank development process. Specifically, existing recommendations for GRM-based analyses would indicate that N = 200–1000 participants are needed to establish stable item parameters (Samejima, 1969; Samejima et al., 1996).

Other published guidelines recommend a minimum of 5–10 individuals per item within an item pool (Bryant & Yarnold, 1995; Everitt, 1975; Gorsuch, 1983). Existing recommendations for DIF analyses indicate the need for a minimal sample size of N = 500 participants (with no fewer than n = 200 participants for each relevant subgroup; Clauser & Hambleton, 1994).

Results

Study Participants

A sample of 218 caregivers of civilians with TBI and 315 caregivers of SMVs with TBI (N = 533) participated in this study (see Table 1 for detailed descriptive data). With regard to injury severity, the majority (57.3%) of civilian caregiver participants in the civilian sample were caring for persons with severe TBI, followed by complicated mild (25.2%) and then moderate (17.4%) TBI. In our military sample, the majority of caregiver participants were caring for SMVs with either mild TBI (21.3%) or undocumented severity (29.8%, most of which were likely also uncomplicated mild TBI given existing rates in the military; DVBIC, 2015). A smaller portion of the military sample were caring for persons with moderate (4.1%), penetrating (6.4%), equivocal (18.7%), or severe (15.6%) TBI. Caregivers of SMVs were more likely to be caring for an individual they perceived to be low functioning (44.6%), relative to their civilian caregiver counterparts (13.8%; χ²(1) = 50.34, p < .001).

Table 1.

Descriptive Information for the TBI-CareQOL study participants

Variable	Caregivers of Civilian TBI (n = 218)	Caregivers of Military-TBI (n = 315)	Combined Caregiver Sample (n = 533)
Age (Years)* M (SD)	51.3 (14.5)	42.2 (11.7)	45.9 (13.7)
Sex (%)*
Female	77.5	94.6	87.6
Male	22.5	5.4	12.4
Ethnicity (%)
Not Hispanic or Latino	90.8	90.5	90.6
Hispanic or Latino	9.2	9.5	9.4
Race (%)*
White	69.7	84.4	78.4
Black/African American	22.0	4.1	11.4
Other	8.3	11.4	10.1
Education (%)*
Less than High School	5.5	2.9	3.9
High School Graduate or	19.7	5.4	11.3
Equivalent
More than High School	74.8	91.8	84.8
Marital Status (%)*
Single, Never Married	18.8	1.3	8.4
Married/Cohabitating	61.9	89.5	78.2
Separated/Divorced	13.3	4.4	8.1
Widowed	4.6	3.5	3.9
Other	1.4	1.3	1.3
Years in Caregiver Role
M (SD)	7.4 (5.4)	6.8 (3.2)	7.1 (4.2)
Relationship to Person with TBI (%)*
Spouse	26.6	77.8	56.8
Parent	38.5	14.9	24.6
Child/Other Family Member	24.3	4.8	12.8
Other (e.g. Friend)	10.6	2.5	5.8
Age of Person with TBI *
M (SD)	43.1 (15.0)	38.9 (9.0)	40.6 (12.0)
Sex of Person with TBI (%)*
Male	72.5	97.8	87.4
Female	27.5	2.2	12.6
MPAI-4 of Person with TBI*
M (SD)	45.2 (16.1)	57.8 (8.9)	53.0 (13.6)
Time since Injury (Years)
M (SD)	9.1 (6.9)	8.6 (3.4)	8.8 (5.4)
TBI Severity** (%)
Uncomplicated Mild	0.0	21.3	12.6
Complicated Mild	25.2	4.1	12.8
Moderate	17.4	4.1	9.6
Penetrating	0.0	6.4	3.7
Equivocal	0.0	18.7	11.1
Severe	57.3	15.6	32.6
Unknown	0.0	29.8***	17.6
Mechanism of Injury (%)
MVA	48.6	40.3	43.7
Falls	16.1	7.0	10.7
Struck by an object or thrown against an object	0.5	23.0	13.7
Gunshot or Assault	19.3	4.2	10.4
Other Accidents (e.g., Bicycle accident, pedestrian struck by motor vehicle)	11.0	3.2	6.4
Sports-related TBI	0.9	1.3	1.1
Other/Unknown	3.6	21.0***	14.0

Note. Entries in the table represent percentage of participants unless otherwise specified;

^*indicates significant group differences: age, t(401.41)=7.74, p <.01, caregivers of individuals with civilian-related TBI were older relative to caregivers of military-related TBI; sex (male, female), more caregivers of individuals with military-related TBI were female, relative to caregivers of civilian-related TBI χ²(1) = 34.64, p < .01; race (White, Black, Other), more caregivers of individuals with civilian-related TBI were Black, relative to caregivers of military-related TBI, χ²(2) = 40.87, p < .0001; education (high school graduate or less, some college, college graduate), caregivers of individuals with civilian-related TBI more likely to have educational attainment of high school or less relative to caregivers of military-related TBI, χ²(2) = 30.16, p < .0001; more military caregivers were spouses of the person they care for compared to civilian caregivers χ²(3) = 141.38, p <.0001; marital status (married/cohabiting vs. all other groups), caregivers of individuals with civilian-related TBI were more likely to be unmarried than caregivers of military-related TBI, χ²(7) = 78.64, p < .0001; gender of the person with injury was more likely to be male in the military group than in the civilian group χ²(1) =74.78, p <.0001; MPAI-4 T score was higher in the SMV group than in civilians (t[491]=11.2; p < .0001; and method of injury was more likely to be other/unknown in the military group χ²(10) = 135.51, p <.0001. Care partners in the military group were younger than patients in the civilian group t(324.22)=3.69, p <.01

^**Documentation of TBI Severity was different for the two groups; for the civilian sample, TBI severity was determined according to TBI model system criteria, whereas for the military sample, TBI severity was determined according to Department of Defense criteria with one exception. Patients with evidence of trauma-related intracranial abnormality and loss of consciousness and posttraumatic amnesia in the mild range, were classified as having a complicated mild TBI (rather than moderate TBI).

^***Documentation of TBI severity was unavailable for the majority of the military sample collected through community outreach by the University of Michigan.

Item Bank Development.

Table 2 outlines the primary findings for the item bank development process. Briefly, EFA analyses supported essential unidimensionality of the pool: the ratio of eigenvalue 1 to eigenvalue 2 was 11.65; eigenvalue 1 accounted for 53.8% of modeled variance, and eigenvalue 2 only accounted for 4.6% of this variance. One item was eliminated due to having a low item-adjusted total score correlation (criterion: r < 0.40), one item was eliminated due to having a low factor loading (criterion: lx < 0.50), six items were eliminated due to local dependence because of their high residual correlations (criterion: r > 0.20), and ten items were eliminated due to local dependence because of their high correlated error modification index values (criterion: MI ≥ 100). Items did not have sparse response cells nor did they exhibit non-monotonicity. Subsequent GRM modeling of the 25 remaining items indicated four items with misfit, resulting in a total of 21 items in the final bank. A final CFA model of these 21 items indicated good overall model fit (Table 3). For the bifactor analyses, in comparing the general factor item slope estimates from a unidimensional vs. multi-dimensional (i.e., one general + two specific factors) model, 19 of 21 items showed no important impact from potential multi-dimensionality; two items had slope absolute value differences of 0.70 and 0.55. However, the general factor had an omega of 0.98 vs. an omega-H of 0.93, indicating negligible impact from multi-dimensionality. In addition, the omega-H values for specific factor models factor 1 and 2 were 0.12 and 0.03, respectively, compared to the corresponding general factor omega values of 0.96 and 0.95, respectively; these findings were also supportive of the ES measure’s essential unidimensionality. Items did not exhibit DIF for any of the tested factors of interest (i.e., age, education, or civilian vs. military status).

Table 2.

Unidimensional Modeling and Analyses

	Unidimensional Modeling								Initial Item Performance		IRT Modeling
Domain	Item pool	EFA E1/E2 ratio (criterion >4)	Percent of variance for E1 (criterion >40)	1-factor CFA loading (criterion <.50)	1-factor CFA residual correlation (criterion >.20)	1-factor CFA modification index (criterion ≥ 100)	Item-adjusted total score correlations (Criterion <.40)	Sparse cells (criterion<10)	Problems with monotonicity	IRT item misfit	DIF	Interim/Final item bank
TBI- CareQOL Emotional Suppression	43 items	Yes	58.1%	1 item	6 items	10 items	1 item	0 items	0 items	4 items	0 items	21 items

Note. CFA = Confirmatory Factor Analysis; EFA = Exploratory Factor Analysis; IRT = Item Response Theory

Table 3.

Final Model Fit and Reliability Estimates for TBI-CareQOL Emotional Suppression

Domain	Item Bank	CFI (criterion >.90)	TLI (criterion >.90)	CFA-based RMSEA (criterion < .15)	SRMR (criterion ≤ .08)	alpha reliability (criterion > .80)	IRT-based RMSEA (criterion < .15)	IRT-based marginal reliability (criterion > .80)
Caregiver Emotional Suppression	21 items	.95	.95	.10	.05	.96	.07	.97

Note. CFI = Comparative Fit Index, TLI = Tucker-Lewis Index, RMSEA = root mean square error of approximation, SRMR = standardized root mean squared residual

The final Emotional Suppression item bank’s estimated item parameters are presented in Table 4. Slopes ranged from 1.68 to 3.33, and thresholds ranged from −2.05 to 2.04. Test information, related to scoring precision, was good for scaled scores between 26 and 74 (see Figure 2 for the scale information function), and IRT-based marginal reliability was 0.97. The maximum number of items (i.e., 12) was used in CAT administrations at roughly ≤ −2.0 SD units and ≥ +2.0 SD units; conversely, CATs tended to use the minimum number of items (i.e., 4) from the item bank from approximately −1.3 to +0.1 SD units (See Figure 3).

Table 4.

TBI-CareQOL Item Parameters for Emotional Suppression

Item	Slope	Threshold 1	Threshold 2	Threshold 3	Threshold 4
I hide my emotions.	2.03	−1.99	−1.01	0.45	1.79
I hide my feelings from the person I care for.	2.03	−1.41	−0.54	0.77	2.04
When I am upset, I hide my feelings from most people.	2.20	−1.74	−0.72	0.32	1.79
I hide my anger.	2.60	−1.73	−0.83	0.46	1.72
I hide my resentment.	1.93	−1.36	−0.66	0.29	1.46
I wait until I am in a private place before getting upset.	2.02	−1.86	−1.07	−0.01	1.35
I keep a smile on my face so others will not know something is wrong.	2.57	−1.83	−1.05	−0.03	1.09
I hide my anger from the people I love.	2.92	−1.52	−0.64	0.29	1.46
I worry about the emotions I show in front of most people.	2.23	−1.36	−0.39	0.53	1.47
I try to seem calm, even when I’m angry.	1.91	−2.05	−1.28	0.03	1.55
I have to think before I speak so I do not express negative feelings in front of most people.	1.68	−1.65	−0.65	0.67	1.93
I hold back my emotions because most people won’t understand what I am going through.	2.22	−1.60	−0.76	0.09	1.16
I hide my emotions from people I am close to.	3.33	−1.43	−0.44	0.39	1.60
I isolate myself emotionally.	2.54	−1.02	−0.33	0.69	1.68
I have to be careful about showing my emotions around the person I care for.	2.27	−1.39	−0.54	0.47	1.39
I sacrifice my own mental health to keep the person I care for happy.	2.08	−0.94	−0.30	0.55	1.48
I wait until I have privacy before I get openly upset.	2.76	−1.54	−0.78	0.11	1.27
I walk away from a situation instead of expressing my feelings.	1.85	−2.01	−1.10	0.49	1.92
I pretend to be happy when I am around most people.	2.15	−1.86	−0.90	−0.05	1.36
If I feel angry, I keep it to myself.	3.02	−1.94	−0.83	0.26	1.54
I don’t confide in most people about my emotions.	2.12	−1.86	−1.05	−0.18	1.37

Note. Items that are indicated in bold were selected for inclusion on the 6-item Emotional Suppression Short Form.

Figure 2. Emotional Suppression Test Information Plot.

In general, we want total information to be ≥ 10.0 and the standard error to be ≤ 0.32 (this provides a reliability of 0.9). This figure shows excellent total information and standard error for Emotional Suppression scale scores between 26 and 74.

Figure 3. Emotional Suppression Number of CAT Items by CAT Theta.

This figure shows the number of CAT items used for different scale scores in standard deviation units: at approximately ≤ −2.0 SD units and ≥ +2.0 SD units the maximum of 12 items from the item bank were used by the CAT; from approximately −1.3 to +0.1 SD units the CAT tended to use the minimum of four items from the item bank.

A six-item short form (SF) was selected from the final item bank using item calibration statistics (e.g., slope, item characteristic curves, item information, and average item difficulty), as well as input on clinical characteristics (e.g., items were selected that represent different ranges of clinical symptoms), by a panel that included experts in TBI, caregivers of persons with TBI and experts in measurement. The score-level reliability for this SF was examined on a measurement continuum from theta = −2.8 (T-score = 22) to +2.8 (T-score = 78). Expected score-level reliability was excellent (≥ .90) for thetas between −1.6 and +1.6, very good or excellent (i.e., ≥ .80) for thetas between −2.4 and +2.0, and good, very good, or excellent (i.e., ≥ .70) for thetas between −2.4 and +2.4.

Scores for the CAT and SF can be calculated using a look-up table (Table 5) or automatically calculated using the different healthmeasures.net administration platforms. Scores are scaled on a T metric (M = 50, SD = 10) relative to other caregivers of persons with TBI. T scores that are ≥ 60 are indicative of clinically significant emotional suppression (relative to other caregivers of persons with TBI); specifically, scores ≥ 60 indicate that the respondent is experiencing emotional suppression that is worse than 84% of their peers. Scores ≥ 70 indicate that the respondent is experiencing emotional suppression that is worse than 95.5% of their peers.

Table 5.

T Score Conversion Table for Emotional Suppression Short Form

TBI-CareQOL Emotional Suppression
Raw Score	T-score	SE *
6	25.51	4.44
7	29.50	3.56
8	32.05	3.26
9	34.13	3.09
10	35.98	2.99
11	37.67	2.92
12	39.27	2.89
13	40.80	2.87
14	42.29	2.86
15	43.77	2.86
16	45.24	2.86
17	46.72	2.86
18	48.20	2.86
19	49.70	2.86
20	51.25	2.87
21	52.87	2.90
22	54.55	2.92
23	56.29	2.94
24	58.07	2.96
25	59.91	2.97
26	61.86	2.99
27	63.94	3.02
28	66.24	3.14
29	68.96	3.47
30	73.08	4.48

^*SE = Standard error

Reliability and Validity Analyses.

Emotional Suppression demonstrated good internal consistency reliability for the full item bank (Table 3) as well as for CAT and SF administrations across both the civilian and SMV caregiver samples (Table 6). In addition, 3-week test-retest reliability exceeded the criterion of 0.70 for CAT and SF assessment in both the civilian and SMV samples (Table 6). Furthermore, there were strong relationships between the scores derived from different administration formats (all rs > 0.96). Emotional Suppression was also free from floor or ceiling effects, and administration time was brief: All forms took less than one minute on average for administration.

Table 6.

Descriptive Data and Reliability Analyses for TBI-CareQOL Emotional Suppression

Administration Form	n	Internal Consistency°	Test-retest reliability *	% of the sample with floor effects	% of the sample with ceiling effects	Mean (SD)	Administration time (sec)	Administration time per item (sec)
			Combined Sample (N = 533)
Short Form	533	0.91	0.91	1.1	3.0	50.10 (9.45)	37.55	6.26
Simulated Computer Adaptive Test	533	0.92	0.89	0.4	1.9	49.95 (9.67)	41.94	7.43
		Caregivers of Civilians (n = 218)
Short Form	218	0.91	0.90	0.9	5.1	47.80 (9.91)	36.90	6.15
Simulated Computer Adaptive Test	218	0.91	0.87	0.0	3.7	47.50 (9.99)	45.12	7.91
		Caregivers of Service Members and Veterans (n = 315)
Short Form	315	0.91	0.88	1.3	1.6	51.70 (8.78)	39.03	6.50
Short Form Simulated Computer Adaptive Test	315	0.92	0.88	0.6	0.6	51.65 (9.09)	37.76	6.80

Note.

^° = internal consistency for static short form measures is reported as Cronbach’s alpha and for computer adaptive tests is reported as average standard error-based reliability (i.e., an item response theory-based reliability estimate);

^*Test-retest reliability is measured by the intraclass correlation coefficient (ICC; criterion ≥ 0.70)

The overall pattern of validity correlations between measures was consistent with our proposed hypotheses (albeit slightly lower in magnitude than anticipated) supporting convergent and discriminant validity (Table 7). Specifically, the strongest correlations were seen between Emotional Suppression and Social Isolation, followed by slightly smaller correlations between Emotional Suppression and both negative aspects of caregiving and mental HRQOL (i.e., magnitudes as hypothesized). Discriminant validity in both samples was supported by negligible to small correlations between Emotional Suppression and measures of Physical Health and positive aspects of caregiving.

Table 7.

Convergent and Discriminant Validity of TBI-CareQOL Emotional Suppression Measure

Emotional Suppression Administration Format	Convergent Validity			Discriminant Validity
	Rand-12 Mental	CAS Burden	PROMIS Social Isolation	Rand-12 Physical	CAS Satisfaction	CAS Ideology	CAS Mastery
Combined Sample (N = 533)
Short Form	0.47**	0.54**	0.59**	0.21**	0.22**	0.14*	0.26**
Computer Adaptive Test	0.47**	0.51**	0.51**	0.21**	0.21**	0.16*	0.28**
Caregivers of Civilians (n = 218)
Short Form	0.38**	0.47**	0.51**	0.25**	0.16*	0.18**	0.26**
Computer Adaptive Test	0.38**	0.47**	0.51**	0.24*	0.17*	0.17*	0.29**
Caregivers of Service Members and Veterans (n = 315)
Short Form	0.47**	0.56**	0.64**	0.14*	0.30**	0.07	0.25**
Computer Adaptive Test	0.46**	0.48**	0.43**	0.14*	0.27**	0.11	0.26**

Note: CAS = Caregiver Appraisal Scale; PROMIS = Patient Reported Outcome Measurement Information System; absolute values for correlations are presented to highlight magnitude

^*p < .05

^**p < .01

Caregivers of high-functioning individuals had significantly lower Emotional Suppression scores compared to caregivers of low-functioning individuals, as hypothesized, supporting known-groups validity. Finally, for the SF administrations of Emotional Suppression, caregivers of persons who were lower functioning were consistently at greater risk for clinically significant emotional suppression than those that were caring for higher-functioning individuals (see Table 8).

Table 8.

Impairment rates for Caregivers of High and Low Functioning Individuals

Emotional Suppression Administration Form	Caregiver of a High Functioning Individual (MPAI-4 < 60)		Caregiver of a Low Functioning Individual (MPAI-4 ≥ 60)
	Mean (SD)	% Impaireda	Mean (SD)	% Impaireda	t	p
Civilian sample	n = 163		n = 26
Short Form	47.01 (9.44)	6.8**	53.12 (8.30)	11.5	3.11	0.002
Computer Adaptive Test	46.54 (9.61)	7.4**	53.30 (8.00)	23.1	3.40	0.001
Service Members and Veterans sample	n = 169		n = 136
Short Form	49.94 (8.51)	8.9*	54.25 (8.03)	17.8	4.50	<.001
Computer Adaptive Test	49.51 (8.95)	11.2	54.15 (8.78)	20.7	4.53	<.001

^aEmotional Suppression T-score > 60

^*p < .05

^**p < .01.

Discussion

The current paper describes the development of a new measure for caregivers of persons with TBI, the TBI-CareQOL Emotional Suppression item bank. This new item bank captures caregiver-reported feelings related to hiding their emotions from others, which is a topic that was verbalized as important by caregivers of persons with TBI who participated in focus groups conducted in preparation for development of the TBI-CareQOL measurement system and noted clinically by many of the investigators. The TBI-CareQOL Emotional Suppression item bank was developed according to established measurement development standards that included classical test theory and item response theory. The final 21-item bank was supported by EFA, CFA, and GRM-based analyses and is devoid of items that are biased for age, education, or caregiver status (civilian vs. SMV). It can be administered as a 21-item long-form (i.e., all 21 items in the bank), 6-item SF, or as a CAT. Previous research has demonstrated that the different administration formats (full item bank, SF, or CAT) are psychometrically comparable (Salsman et al., 2019; Stucky, Huang, & Edelen, 2016) and there is evidence to support that the generally outperforms similar static measures with substantially more items (Pilkinois et al., 2011). Psychometric performance is also comparable for different administration modes including interviewer-administration, self-administration at home (computer), self-administration in-person (computer), paper administration, personal digital assistant, and interactive voice response administration (Bjorner et al., 2014; Kisala et al., 2019; Magnus et al., 2016; Wang, Chen, Usinger, & Reeve, 2017). Practical reasons may dictate which administration format and mode might be most feasible for any given study (e.g., burden, hardware/internet requirements). CAT and SF administration offers the advantage of brevity (typically 4 to 12 items in length), better precision and lower standard errors than more traditional static measures, even for administration formats of equal length (Lai et al., 2011). This brevity facilitates rapid serial administration of several measures in a short period of time ultimately reducing response burden which can be especially problematic in a population that is already overwhelmed by their caregiver roles and responsibilities.

There is also considerable evidence to support the standard-meeting psychometric properties of the new TBI-CareQOL Emotional Suppression measure, as obtained in use with both caregivers of SMVs with TBI and caregivers of civilians with TBI. With regard to reliability, for both groups (SMVs and civilians): 1) score-level reliability was supported; 2) internal consistency was excellent, regardless of the administration format (i.e., all αs ≥ .91); and 3) test-retest reliability ranged from good to excellent across administration formats (all ICC ≥ 0.87). In addition, none of the Emotional Suppression’s administration formats exhibited floor or ceiling effects.

The pattern of correlations observed between Emotional Suppression and other measures of HRQOL and caregiver functioning provide us with preliminary support for convergent and discriminant validity. Our hypotheses that Emotional Suppression would be moderately correlated with Social Isolation and emotional HRQOL, and negative aspects of caregiving was supported by the data. The strongest relationship was seen between Emotional Suppression and Social Isolation (in the moderate range), followed by slightly smaller correlations between Emotional Suppression and negative aspects of caregiving (i.e., burden) and between Emotional Suppression and mental HRQOL. The correlations between Emotional Suppression and these measures of social isolation and emotional HRQOL and negative caregiving were lower than the anticipated magnitude. However, this could be due to the fact that we did not include a criterion measure of emotional suppression in this study. This is a limitation of the study and future work to marshal additional evidence of construct validity could utilize criteria measurement tools that focus on emotional suppression.

Similarly, our hypotheses that Emotional Suppression would not be related to either physical health or positive aspects of caregiving were supported by negative aspects of caregiving and Mental HRQOL. Negligible to small correlations were found between Emotional Suppression and measures of physical health and between positive aspects of caregiving. This supports discriminant validity. It is important to note that, while the pattern of correlations was consistent with our proposed validity hypotheses, the magnitude of the correlation between Emotional Suppression and its closest comparator (i.e., Social Isolation) was slightly smaller than hypothesized. We expected the strength of this relationship to be demonstrated by an Emotional Suppression vs. Social Isolation r of > 0.60; observed rs ranged from 0.43 to 0.64 across the different subsamples of caregivers and/or administration formats that were examined. This finding is consistent with literature that would suggest that it is not the magnitude of the correlation that is most important, but rather that scores for discriminant measures should be noticeably lower than scores on convergent measure (Hubley, 2014). Thus, we still interpret these findings as support for convergent validity, and we conclude that, while these two constructs are indeed overlapping, they are more distinct than we had initially hypothesized (Note: We are currently unaware of any published literature that would support or refute this assumption.) It may be that caregivers are involved in social activities, but still feel the need to suppress emotions when in these social settings in order to be socially appropriate or not make others feel uncomfortable. This suppression behavior then allows them to engage in some social events that may be rewarding but limits the amount of social support they actually receive because they feel the need to keep their emotions in check and not share their feelings with others. Regardless, relationships between Emotional Suppression and caregiver burden and between Emotional Suppression and emotional distress are consistent with other caregiver research in persons with TBI (Davis et al., 2009; Hanks et al., 2007; Sander et al., 1997).

TBI-CareQOL Emotional Suppression was also able to differentiate between caregivers of high-functioning individuals versus vs. low-functioning individuals with TBI: Caregivers of high functioning individuals consistently reported less emotional suppression than those of low-functioning individuals, regardless of caregiver group or test administration format. These findings provide preliminary support for the known-groups validity for this new measure. As such, this measure may be used to compare groups in other samples and to test hypotheses such as whether a group receiving psychological treatment for emotional suppression has better scores than a control group.

Finally, while caregivers of SMVs and civilians appear to have different scores on TBI-CareQOL Emotional Suppression, all of the psychometric analyses that we have conducted all suggest that these measures are not biased against one group or the other (i.e., items are free of DIF), and that the measure performs equally well in both subsamples (in terms of reliability, convergent/discriminant validity, and known groups validity). These group differences in scores are not especially surprising and may be due to a number of important sample differences which include different inclusion criteria (thee SMV sample included mild TBI, whereas the civilian group did not), as well as several differences on demographic variables (see Table 1) that include differences in caregiver age (the SMV group is younger), relationship to the person with TBI (the SMV group is more likely to be spouses), and the overall functional status of the person with TBI (the SMV group were more likely to be caring for an individuals that were perceived as being more impaired). There are also differences in the recruitment sources for participants (civilian recruitment was typically hospital-based, whereas SMVs were recruited through both hospital and community-based efforts that may contribute to group differences. While findings support the reliability and validity of this new caregiver-specific HRQOL measure, there are several study limitations that warrant acknowledgement. First, while this study focused on both caregivers of SMVs and caregivers of civilians (with reported findings similar across these two samples), it is important to recognize that these groups were different in several different ways: Caregivers of SMVs with TBI were younger, caring for an individual that was younger, were more likely to be women, and more likely to be the spouse of the person with TBI, than caregivers of civilians with TBI. In addition, the SMV sample included caregivers of persons with uncomplicated mild TBI, and thus had broader representation across the TBI severity spectrum than the civilian sample. Additionally, the MPAI-4 used to determine known groups validity is a caregiver proxy measure of functional impairment in the person with TBI. Caregivers may have different perceptions of impairment and the pre-injury status of the person with TBI could strongly influence proxy ratings. These group differences may explain some of the small differences in the overall psychometric performance of this new measure, and more research is needed to directly compare these two groups (especially with regard to TBI severity, which was missing for a large portion of the SMV sample, as well as objective ratings of disability for the person with TBI). Also, regardless of group (SMV vs. civilian), caregivers were predominantly female, and thus generalizability to male caregivers (especially within SMV samples) needs to be established. Future work is also needed to establish the responsiveness of this measure to change over time, particularly treatment-related change, among these caregivers. Future work could also consider a more nuanced examination of the impact of different types and combinations of care assistance (e.g., physical assistance, emotional assistance, financial assistance) on caregiver outcomes.

Taken together, the TBI-CareQOL Emotional Suppression item bank provides a psychometrically robust assessment of emotional suppression in caregivers of SMVs and civilians with TBI. The measure provides a brief assessment (less than one minute for administration of either SF or CAT measure versions) that can be used to identify caregivers who may be at risk for negative effects of use of emotional suppression. It should be noted that emotional suppression is a single type of emotion-focused strategy that may be maladaptive, but there are other similar strategies, including avoidance of the experience of emotions and denial of unpleasant emotions. These are not covered by the emotional suppression item bank and may warrant assessment. Nonetheless, the Emotional Suppression item bank may be used to identify caregivers who are using one form of maladaptive emotional regulation and to refer them for treatments, guidance, and education about developing healthier coping strategies. This has the potential for downstream effects that can improve HRQOL outcomes for caregivers across the health spectrum.

Impact and Implications Statement:

• Research suggests that caregivers that use coping strategies focused on denying or avoiding thoughts and emotions experience higher levels of caregiver burden.

• This new measure, TBI-CareQOL Emotional Suppression, is a new patient-reported outcome measure designed to identify caregivers of persons with traumatic brain injury that might benefit from clinical interventions that foster more positive coping strategies.

List of Abbreviations:

CAS

Caregiver Appraisal Scale

CAT

Computer Adaptive Test

CFA

Confirmatory Factor Analysis

CFI

Comparative Fit Index

DIF

Differential Item Functioning

DoD

Department of Defense

DVBIC

Defense and Veterans Brain Injury Center

EFA

Exploratory Factor Analysis

GRM

Graded Response Model

HRQOL

Health-Related Quality of Life

IRT

Item Response Theory

MHC

Mental Health Composite

MPAI-4

Mayo-Portland Adaptability Inventory-4

PHC

Physical Health Composite

PRO

Patient Reported Outcomes

PROMIS

Patient Reported Outcomes Measurement Information System

RMSEA

Root Mean Square Error of Approximation

SF

Short Form

SMV

Service Member/Veteran

TBI

Traumatic Brain Injury

TBI-CareQOL

Traumatic Brain Injury Caregiver Quality of Life

TLI

Tucker-Lewis Index

WRNMMC

Walter Reed National Military Medical Center