A psychometric evaluation of a new social subscale for the Common Misconceptions about Traumatic Brain Injury (CM-TBI) questionnaire: toward the CM-TBI-II

ABSTRACT Objective Existing TBI misconception measures are critiqued for failing to measure postinjury social experiences. This study developed a social subscale for the Common Misconceptions about TBI (CM-TBI) questionnaire for use in the general public. Methods Seven experts independently review items drawn from the literature. Shortlisted items were administered online to 158 adults (aged ≥18 years; 51% postschool educated; 60% no TBI experience), the CM-TBI, and a measure of construct validity (a published TBI-adaptation of the Community Attitudes Towards the Mentally Ill; CAMI-TBI). One week later, the new items were redeployed (n = 46). Results Expert review and iterative correlations identified a 10-item social subscale (internal consistency, test-retest reliability, α’s>.80). When added to the CM-TBI (ie. CM-TBI-II), the internal consistency was .71. The social subscale was significantly correlated with CAMI-TBI measures (p’s <.05, r’s > .3). There was no significant difference on the social subscale for education subgroups (school vs post-school, p = 0.056) or previous TBI experience; but there was a difference for the CM-TBI-II (post-school>school; Cohen’s d = 7.83, large effect). Conclusion This study found strong preliminary psychometric support for a new social subscale, administered as the CM-TBI-II. This subscale shows promise as a measure of misconceptions about social functioning post-TBI. The CM-TBI-II could support evaluations of programs aiming to improve social engagement and community participation for people with TBI.


Introduction
Traumatic brain injuries (TBI) have a high rate of incidence globally (1) and occur after an external blow or sudden jolt to the head, commonly from motor vehicle accidents or falls (2). TBI is classified as mild through severe depending on clinical indicators such as the duration of loss of consciousness or posttraumatic amnesia (1). Although TBI outcomes are heterogeneous (1), people's social experiences can be significantly negatively affected following mild through severe injury (3).
There are many ways in which the social experiences of an injured person can be negatively affected (2,4). For example, the ability to use or read social cues can be reduced after severe injury, and new social challenges can be encountered because of public misconceptions of about TBI. Persons with mild through severe TBI describe ongoing social isolation, disrupted, or altered relationships and sense of self, and negative evaluations by others (5)(6)(7)(8). There is growing recognition of the importance of improving social outcomes after TBI, and this could include a role for determining if the general public does in fact hold misconceptions about the social effects of TBI.
According to general (9) and TBI-specific models of functioning (4,10,11), when an injured person returns to the community after receiving medical care, their experience is shaped by their environment. This includes their access to social supports, and the availability of social interaction partners (such as employers, spouses, neighbors). The injured person may be able to access or receive social support from proximal (e.g., family) and distal sources (e.g., community, employer, service providers), and the nature of the social interactions with these sources will depend on the recovery stage (12). Further, while different terminology is used to describe social experiences after TBI (12), it is clear that community views about TBI and the capacities of the injured person (e.g., degree of social cognition impairments) can interact with each other. For example, residual TBI effects after more severe TBI, such as decreased emotion recognition, can be 'invisible' (13)(14)(15), 'non-visible' (16), or misunderstood by community members, leading to disadvantage and marginalization of the injured person (4,(17)(18)(19).
Prior research has investigated community-held views about TBI and if these could affect social interactions with people with TBI or reduce the availability of social support experienced by them. Experimental studies show that community members are more likely to ascribe negative attributes to a person with a brain injury than they are to a person with a different (non-brain) injury (19). In a study of 323 students and community members, community members were more likely than students to associate negative attributes (such as untrustworthiness) with brain injury (20). If such views function to worsen social interactions and TBI outcomes as proposed (4), there is the potential for change through community education.
The idea of improving outcomes for persons with TBI through increased community awareness about this injury is not new (21); however, an expanded approach that considers public awareness of the social effects of TBI is now needed. Past studies have focused on public TBI education to improve injury identification and management, and increase awareness of prevention and rehabilitation strategies (22,23). This has included studying the knowledge gaps, including in high injury-risk groups and if misconceptions can be changed (4). Table 1 shows a list of existing tools used for this purpose and their main features. The existing measures are often specific in their focus and target populations, highlighting some limitations of these tools. Existing TBI misconception tools have been strongly critiqued on several grounds (4) including that they are not conceptually grounded, there are item-scaling limitations, and some lack psychometric evaluations, especially during their development. Most importantly, these tools have been critiqued because they overlook social factors, such as injury visibility and community attitudes toward persons with TBI (29,32,33).
Given that there is no existing TBI misconception measure with questions about social attitudes toward persons with TBI (33), this study sought to address this gap. The purpose of this study was to develop and psychometrically evaluate an item set to be added to the most established, well-researched and widely used measure of TBI misconceptions: the Common Misconceptions about TBI questionnaire (CM-TBI). Guided by best practice recommendations for test development (34) a multi-stage process was employed, including item identification via expert (health professional) consultation and literature review, and item selection with expert input (phase one). Phase two evaluated the new items when administered as an extension to CM-TBI. The phase two aim was to determine the psychometric properties of the new items and the extended test (CM-TBI-II), including through investigations of internal consistency, test retestreliability, and construct validity.

Participants
16 TBI experts across several Australian states were approached for phase one. Experts were defined as health professionals who were published in TBI. This pool included people in multidisciplinary fields, including neuropsychology, occupational therapy, and neuroscience. The experts were individually approached by e-mail in September 2021 and asked to complete an online survey [Qualtrics™, Provo, UT]. Seven experts evaluated the new items (44% response rate). This number satisfies the best practice recommendation that an initial item pool should be reviewed by five to seven experts (34).
Initially, two hundred and nineteen people commenced phase two. These participants were recruited by word of mouth through the researchers' network, and via electronic distribution of project information to staff and students at a large metropolitan university. Project information was also disseminated to a research participation pool, and via a community volunteer database. Eligible participants were aged ≥18 years. After screening for compliance with the age requirement, and screening for 'invalid' responses (see screening process below), 158 participants completed session one of phase two (T1), and 46 participants returned for session two (T2); the retest survey, administered one week later.

Shortened Common Misconceptions about TBI (CM-TBI) and CM-TBI-II
The CM-TBI is a 40-item measure of TBI knowledge gaps with good psychometric properties (e.g (26,35). The CM-TBI records responses on a 4-point scale (true, probably true, probably false, false). Several variants have been used to study TBI misconceptions in various populations (25,(36)(37)(38)(39)(40)(41), including educational professionals (42) and probation board members (35). A 20-item (shortened) version was developed using principal components analysis, and this identified four subscales: recovery, sequelae, hidden injury, and insight (26). This shortened variant assessed pediatric TBI and used a modified (5-point) Likert response scale (1 = strongly disagree to 5 = strongly agree) with a 'don't know' mid-point. An example item is: It is good advice to remain completely inactive during recovery from a brain injury. The 20-item version was subsequently used in a study of adult TBI (39). For this study, we used the 20-item adult version (39), with a new mid-point ('neither agree nor disagree') to improve scale continuity (33). Items were also added to the measure, based on phase one. The added 13 items assessed beliefs about the social experience of persons with TBI, including attitudes about social functioning or behavior, and broader environmental factors (e.g., opportunity for community participation). After score reversals, a total agreement score was calculated (possible score range: 20-100), with higher scores indicating stronger misconceptions. Thus, the possible score range for the new items was 13 to 65, and for the expanded measure (i.e., new plus existing items [CM-TBI-II]), the possible score range was 33 to 165.

Community Attitudes toward the Mentally Ill (CAMI) -TBI modification (CAMI-TBI)
The CAMI is a well-established measure of community attitudes toward individuals with a mental illness (43). A previously modified 40-item version of the CAMI was employed in this study as a measure of validity for the new items (44). In this version the term 'mental illness' is replaced with 'brain injury' (i.e., CAMI-TBI) (44). The CAMI-TBI has been used to measure outcomes from a school-based intervention for reducing stigmatization of acquired brain injury (45). An example CAMI-TBI item is: Individuals with a brain injury are far less of a danger than most people suppose. The same 5-point agreement scale was used as for the CM-TBI-II. The CAMI-TBI has a multi-component structure (10 items per component), and in its original form, the subscales have satisfactory reliability [10]. The CAMI-TBI is scored by re-scaling the reversed items, calculating item ratings, and summing the item ratings for each of the four component scores (authoritarianism, social restrictiveness, benevolence and community ideology) (44). Higher authoritarian and social restrictiveness scores are indicative of more unfavorable attitudes toward persons with TBI, whereas higher benevolence and community ideology signify more inclusive and supportive attitudes. The CAMI-TBI items from this study are available on request.

Previous experience with TBI
Previous experience with TBI was assessed with a single multiple-choice item focused on the most common form of injury (mild TBI). We asked: Do you have any previous experience with a mild traumatic brain injury? The three response options were converted to a binary score to create a group with (TBI-Experience) or without previous experience (TBI-No Experience). The group with experience was a composite of those who responded that they or someone they knew had experienced TBI.

Validity measures: Marlowe-Crowne Social Desirability Scale and Instructional
The 10-item Marlowe-Crowne Social Desirability scale (M-C 1 (10)) was administered as a control for a potential confound (46). This scale has a binary response format (true = 1, false = 2) and with reverse scoring of items with a 'true' response (47). Total scores on the M-C 1(10) can range from 10-20, with higher scores indicating higher levels of socially desirable responding (46,47). A three-item instructional manipulation check (IMC) was deployed at T1, and one IMC at T2. IMCs can provide an indication of invalidity (e.g., inattentiveness) in online surveys (48,49). IMCs tell the participant which response to select (e.g., 'please select agree'), and are scored as correct or incorrect. A lower total IMC score indicates a higher risk of invalidity. At T1 an IMC cut off at least 50% correct was applied to identify valid protocols (low inattentiveness risk); at T2 the IMC cut off was 100% correct.

Procedure
This research was part of a larger study. Applications for ethical (identification number: 4414) and occupational health and safety clearance (identification number: 1782) were approved by the Queensland University of Technology (QUT). In phase one, the experts were told that the study aimed to modify the CM-TBI to include items about social functioning and community attitudes toward persons with TBI. They were shown the original CM-TBI and twenty-four pilot items identified by the authors from a review of literature examining the social experiences of persons with TBI and public misconceptions about TBI that might impact those social experiences (6,7,20,44,50), plus related social attitude scales, such as the TBI adaptation of the Community Attitudes toward the Mentally Ill (44). The pilot items were modelled to fit with existing CM-TBI items (e.g., by matching stems, where possible, and other key referents (e.g., injury terminology)). The pilot items were then evaluated by the experts (value-add, relevance, understandable, ambiguity) and rated (top 3 'worst' and 'best' items, respectively), and they provided input at a granular-(e.g., is the item unambiguous?) and global-level (e.g., is a revised scale needed?). Experts' views were also sought about the community attitudes toward TBI and any additional recommended items, and whether the pilot items provided adequate coverage of this aspect.
In phase two, the study was advertised with materials presented online (Qualtrics™). Survey access was controlled via an opt-in method (activating a link, scanning a quick response (QR) code). A full consent form was shown including information regarding risks and benefits of participation, privacy, and complaint procedures. A forced response was required to proceed (click 'agree') or exit ('disagree'). Conditional logic directed consenting participants through the remaining questionnaire (non-forced responses). Participants generated a unique identifier to link anonymized responses, then answered demographic questions, followed by the CM-TBI-II, and the CAMI-TBI (with one embedded IMC). The CM-TBI-II was administered in two fixed blocks, each with one embedded IMC, plus the new items (block 1), or the original CM-TBI items (block 2). One week later an automated e-mail with a T2 survey link comprising block 1 items was sent to participants who requested it. Exiting participants either quit the study immediately, or after they entered their contact details into a separate database held for administrative purposes (e.g., token distribution). Two tokens of appreciation were offered: a prize draw entry (one chance to win one of two $AUD50 gift cards) or 1% academic credit, if eligible. 78 participants received a study summary and two participants received prizes (drawn May 2022).

Data analysis
The data were transferred for analysis into the Statistical Package for the Social Sciences (IBM SPSS Statistics ™ version 28). The phase one results (item-and scale-evaluations) were examined using descriptive statistics, with counts employed in item decisions (e.g., retention, revision, removal). The qualitative results were used for item refinement. The phase two data was screened for missing results (>10%) and patterns (Little's MCAR, p's < .05), and invalid attempts. The social desirability scores on the M-C 1 (10) were examined with reference to conventional standards (48,49) and did not indicate socially desirable responding. Intercorrelations between new item pairs were inspected to explore interrelationships (51). Based on precedent (34), potentially problematic items were flagged using an a priori standard (i.e., bivariate intercorrelation < .3). Internal consistency (Cronbach's alpha) was explored for the CM-TBI, new item set, and the CM-TBI-II. A oneweek test-retest intraclass correlation coefficient (ICC) was generated for the new items (52). Validity was examined through the CM-TBI-II and CAMI-TBI intercorrelations. Independent-samples t-tests examined differences in CM-TBI-II scores for people based on previous education and experience of TBI, respectively, and effect size was calculated with Cohen's d. A p-value of .05 was used for statistical significance, unless otherwise stated.

Phase 1: Expert review
The expert feedback was tabulated and descriptive statistics calculated where appropriate. Thirteen items were retained because six or more experts deemed them valueadding or relevant to community attitudes. For the eleven items that were removed, three or more experts did not rate them highly (low value-add or relevance), or they received at least two votes as the 'worst' item from the pool. A selection of the qualitative feedback is shown in Table 2. This table shows how

Phase 2: Community sample
The flowchart in Figure 1 explains the participant selection, and Table 3 displays the sample characteristics (N = 158). The average age of the participants was approximately 29 years (SD = 15). Seventy-eight percent of the sample (n = 124) identified as female and 41% self-reported previous experience with TBI (either personally or indirectly, n = 64). Approximately half of the sample had completed their school education (n = 77) with the other half having also completed a post-school qualification (n = 81).

CM-TBI-II-Social subscale: item reduction
Intercorrelations and reliability analyses were performed on the new CM-TBI-II social subscale items. This information was used to identify items with weak inter-item associations (<. 3) (34) and/or whether their removal would increase the scale alpha. An iterative process was used to remove items until the values in the intercorrelation matrix approximated or exceeded .3 and the reliability analysis showed that the removal of further items would not significantly increase the scale alpha. Three items were excluded in this process. The intercorrelations for the remaining 10 CM-TBI-II items are shown in Table 4 (lower diagonal). These correlations ranged from .13 I would employ a person with a brain injury. I would not trust a person with a brain injury to take care of my finances.
Items retained unchanged.
The social experience post-TBI can include decreased social status, and increased social deprivation and isolation (4,53) Experts recommended coverage of: 'separation from peers due to injury;' 'impact on friendships and activities with friends' People with a brain injury are generally negative and anti-social. I would be hesitant to befriend a person with a brain injury.
Items retained, with rewording for one item: 'It is common for a person with a brain injury to show anti-social behaviors.' Note. Experts also suggested further wording alterations for several phase one items to improve the match with existing CM-TBI items (replacing 'individual' with 'people' or 'person' with TBI). This change was implemented. The tabulated expert recommendations were elicited from an item about the key social considerations post TBI.
(small) to .55 (medium (54)), and all were statistically significant bar one correlation. The final set of items had an alpha of .82 at T1 and .87 at T2 (Table 5).  Table 5).

CM-TBI-II validity
The validity of the CM-TBI-II social subscale was examined through intercorrelations with existing CM-TBI subscales, and the CAMI-TBI subscales, respectively. Table 4 shows selected correlations for the CM-TBI-II social subscale and CAMI-TBI subscales (upper diagonal). The new CM-TBI-II social subscale was significantly correlated (p < 0.05) with one existing CM-TBI subscale (i.e.,

Misconceptions in the community and the effect of individual differences on item and summary scores
Measures of central tendency and dispersion for the average agreement on CM-TBI-II and CAMI-TBI subscales are shown in Table 5 (full sample and grouped by education). A series of repeated measures t-tests compared the average score on the CM-TBI-II social subscale (M = 3.85, SD = .50), paired with an existing subscale (M = 3.17 to 4.17). The mean score for the social subscale was significantly higher than the sequalae and insight subscales (p's<.001), and significantly lower than the recovery (p=0.024) and insight subscales (p < 0.001).
Independent-samples t-tests were used to examine educationor experience-related group differences on the CM-TBI-II. There was no significant difference based on previous or no previous TBI experience for any item, subscale, or total score (all p's > 0.05). There were education group differences, but this depended on the score type (item, subscale or total), and when observed they were in the same direction (post-school > school). Specifically, education-group differences were found at the itemlevel for two new (social) items (Item 1 and 7) and four original items (Items 2, 7, 11 and 14; all ps < .05, ds >.8, large effects (54)).  Note. CM-TBI = Common Misconceptions about Traumatic Brain Injury. CM-TBI-II = Extension of the CM-TBI that includes a 10-item social subscale. CAMI-TBI = Community Attitudes toward Mental Illness (TBI adaptation). CMI-TBI-II Social subscale items: 1: It is common for people with a brain injury to be easily identified from those without a brain injury † ; 2: It is usually best to avoid people with brain injuries † ; 3: People with a brain injury are usually unable to handle important responsibilities † ; 4: I would be hesitant to befriend a person with a brain injury † ; 5: People with a brain injury can be productive members of society; 6: It is usually difficult to hold a conversation with a person with a brain injury † ; 7: People with a brain injury are usually unable to properly think for themselves † ; 8: I would not trust a person with a brain injury to take care of my finances † ; 9 : I would employ a person with a brain injury; 10: It is common for a person with a brain injury to show anti-social behaviours † . † = Reversed items. Subscales: 1: CM-TBI-II-Social; 2: CAMI-TBI-Authoritarian; 3: CAMI-TBI-Benevolence; 4: CAMI-TBI-Social restrictiveness; 5: CAMI-TBI Community mental health ideology. * p < 0.01 ** p < 0.001.

Discussion
This study responds to a call from the field for a revised measure of TBI misconceptions that considers the social experience post TBI (33). The capacity to evaluate and change misconceptions about the social experience of persons with TBI has arguably been limited by the absence of a suitable tool for this purpose (12). Consistent with this idea (12), none of our experts knew of a measure for this purpose, but all recognized this need as very important. Because existing TBI misconceptions measures have also been critiqued for insufficient psychometric evaluation when under development (33), the current study aimed to build and psychometrically evaluate a 'social extension' for an existing misconceptions measure with known properties (i.e., the CM-TBI).
To establish a basis for comparison, this study first replicated the psychometric properties for the existing CM-TBI subscales. Consistent with past reports, this study found acceptable (>.70) internal consistency for some CM-TBI scores (total score, recovery subscale (26,35)); but not for others (Insight (35)). O'Rourke et al (35). found that the sequalae subscale was internally consistent (α = .81), whereas this result was not replicated (α = .69). The new social subscale had an internal consistency of .81, and when added to the CM-TBIto create the CM-TBI-II -the alpha for the CM-TBI-II was acceptable (above .70) (34). Importantly, the new social subscale also had acceptable one-week test-retest reliability (>.80) signifying stable measurement. Taken together, this investigation suggests that the existing CM-TBI subscales could be further developed (for example via item addition for some subscales), and the new scale performed relatively well.
The relationship between attitudes toward social aspects of TBI, and existing TBI misconceptions was explored. If the new scale is a valid measure, it should be related to existing misconceptions. These patterns were indeed evident; for example, there was a significant positive correlation between the social and recovery subscales, suggesting that with increased knowledge of the social effects of injury, there is increased knowledge of general TBI recovery. However, social knowledge was not significantly correlated with other misconceptions (insight or hidden injury), and further, it was significantly negatively correlated with an understanding of TBI sequalae. This significant negative correlation could indicate that prosocial attitudes toward a person with TBI can be high, even when symptoms are not fully comprehended (33). Overall, while these CM-TBI-II subscale analyses require cautious interpretation due to the low internal consistency of the existing subscales, it is promising that the new social subscale was significantly associated with the most reliable of the existing subscales. A suggestion for future research is to trial the new subscale in the full 40-item CMI-TBI, or in a modification of the shortened version with added or improved items for the two weakest CM-TBI subscales (hidden injury, insight).
The pattern of correlations between the CM-TBI-II social subscale and CAMI-TBI subscales was explored to test for construct validity. Specifically, this determined if the new items were related to conceptually-similar items from a measure of community attitudes toward people with TBI (44). Consistent with expectations, there were significant, moderate, positive correlations between the new social subscale and two of the four CAMI-TBI subscales (Benevolence and Community Ideology) and significant negative correlations with the others (Authoritarianism and Social Restrictiveness). The negative correlations suggest that stronger prosocial TBI attitudes are associated with weaker attitudes about the control of, or social restrictions for, persons with TBI (i.e., lower stigma and discrimination). The positive correlations suggest that as prosocial attitudes increase so too do attitudes that reflect a humanistic view of TBI and a valuing of community-integrated supports (i.e., higher social acceptance and understanding). This suggests that prosocial attitudes, such as a willingness to engage with a person with TBI, extends the support for community-integrated services and a more benevolent injury appraisal. Overall, these results support the hypothesized relations between the new subscale and an established measure, thus displaying convergent validity.
As a further validation, this study tested for differences in TBI misconceptions based on education (school versus postschool) or TBI experience (previous vs no previous TBI experience) since studies have shown that knowledge can be influenced by such factors (4). Contrary to expectations, this study did not find a statistical difference in TBI misconceptions due to experience; but it did find an effect for education. The experience effect is inconsistent with a prior study that showed that TBI misconceptions were reduced by previous TBI experience (35). However, on closer inspection of that prior study, the experience effect was quite limited (i.e., effects evident on two of four subscales [sequalae, hidden injury], and one of two measures ('having knowledge of someone with a brain injury' but not having 'experience of working with a person with a brain injury,' p.1121) (35)). Taken together, the role of TBI experience on misconceptions -including those measured via existing subscales -is not clear, potentially because of different measures.
Consistent with the education hypothesis, this study found an effect of general education on the CM-TBI-II total score (Cohen's d = 7.83, large effect). This is consistent with past studies showing that higher levels of education are associated with fewer TBI misconceptions (4,38,56). However, an education effect was not found for any of the subscales (albeit that, the difference approached significance for the new subscale, p= 0.056). This finding could temper conclusions about subscale validity or interpretability and requires replication. If the result stands, it suggests that formal education is unlikely to provide an avenue for supporting people to hold more accurate views about social functioning post-TBI, and that other strategies will be needed.
This study has several limitations. Online surveys carry a risk of bias, as do attitude surveys (e.g., socially desirable responding (48,49)). To address this issue, this study employed data quality checks and excluded cases for reasons such as inattentive responding. Despite this, the findings may carry a risk of bias. Second, these findings have limited generalizability because the sample was small (especially at retest), unrepresentative of the community (comprised of mostly women volunteers, with almost half of the sample holding a post-school qualification), and some effects were small. In a sample that is more representative of the general public, it is speculated that additional misconceptions would be revealed. Third, while the psychometrics for the new scale were promising, the comparative measures and their subscales (the gold standard) have known weaknesses, some of which were replicated in this study. The CAMI-TBI has only been used in a handful of studies, and the validity and interpretability of CM-TBI-II scores (total and subscale) requires further investigation in a larger study with structural analyses (e.g., factor analysis). Until this occurs, it cannot be concluded that the new subscale taps into a unique dimension or that it can be interpreted alone. Future studies can build on this research to address these limitations, such as employing alternate measures of a range of variables, including TBI experience. Future research should also investigate community attitudes about social functioning in people with mild versus severe TBI.
The social experience of persons with TBI can be debilitating and can reduce community participation (6,7). Modelling suggests that many factors contribute to this problem, including the role played by the community via beliefrs or attitudes they may hold (4,10). This study filled a gap in the literature through the development of a social extension for a wellestablished measure of community misconceptions about TBI. If this measure is deployed in future research it could assist in understanding community views about the social engagement of persons with TBI. Further, it could prove a useful tool for the evaluation of programs aimed at improving the social experience and community participation of persons with TBI. While additional testing is needed for the social subscale developed in this study, including via independent replication and extension, the findings strongly suggest that this new subscale has promising psychometric properties.