Development and validation of the Trauma-Related Cognitions Scale

Cognitive theories suggest the manner in which individuals process trauma-related information influences posttraumatic sequelae. Interpretations about trauma can be maladaptive and lead to cognitive distortions implicated in the development of posttraumatic stress disorder (PTSD) through the processes of overaccommodation and assimilation. Alternatively, adaptive interpretations about trauma through the process of accommodation can lead to post-trauma resilience and recovery. The Trauma-Related Cognitions Scale (TRCS) provides a measure of beliefs associated with these cognitive processes. The TRCS was developed over the course of four phases. During Phase 1, 94 items derived from previously validated trauma cognition/beliefs measures were aggregated with 40 items developed by the authors. Phase 2 investigated the TRCS factor structure by fitting exploratory factor analysis (EFA) models to data from a non-clinical sample, resulting in a reduced 69-item TRCS representing four factors: the three theoretical cognitive processes of overaccommodation, assimilation, and accommodation, and an additional optimism factor. Phases 3 and 4 fit confirmatory factor analysis (CFA) models of the 69-item TRCS in a new non-clinical and a clinical sample, respectively, and further validation analyses were conducted. Initial evidence suggests the TRCS is a valid and reliable measure of trauma beliefs. Continued validation can determine its utility in both research and clinical contexts.


Introduction
Most individuals experience at least one traumatic event in their lifetime [1]. Although posttraumatic stress symptoms are common reactions to traumatic events, the typical outcomes following these events are recovery or resilience [2,3]. In fact, a large study with adults in the U.S. (N = 2,953) estimated that posttraumatic stress disorder (PTSD) affects less than 10% of adults exposed to traumatic events [1]. Given that only a minority of trauma-exposed individuals go on to develop PTSD, continued empirical identification of factors associated with the development and/or maintenance of PTSD symptoms is warranted.
Prominent social cognitive and information-processing theories have historically recognized that the manner in which individuals process trauma-related information influences trauma-related sequelae [4][5][6][7]. Cognitive processes that have been implicated in the development of posttraumatic stress symptoms include over-accommodation and assimilation [7,8], while accommodation has been associated with recovery [9]. Accommodation involves the alteration of cognitions to incorporate new post-trauma information, and is typically associated with balanced cognitions about the traumatic event, the self, and the world (e.g., "I was hurt by another person, but that does not mean everyone is bad."). By contrast, over-accommodation processes result in an alteration of post-trauma cognitions that are more extreme and broader in nature, such as overgeneralizing (e.g., "People are inherently evil") or catastrophizing (e.g. "My life will never be the same"). Assimilation involves altering new information to maintain and reinforce pre-existing cognitions (e.g., "It is my fault this happened."). These theories suggest that erroneous or extreme interpretations of the causes and/or consequences of the traumatic event interfere with recovery. Moreover, individuals may develop problematic patterns of interpreting new information based on cognitive distortions developed posttrauma. Distorted thinking may lead to avoidant behavioral and cognitive coping responses intended to reduce the sense of threat (e.g., avoiding crowds due to fear of unknown danger, constantly scanning one's environment for threats even in relatively safe locations). However, these strategies prevent change in event-related interpretations and instead maintain posttraumatic stress symptoms such as avoidance, irritability from physiological arousal, and distress caused by intrusive thoughts and memories [10,11]. Trauma-related changes in cognitions have become explicitly recognized as diagnostic criteria important to the development and trajectory of PTSD within the most recent edition of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) [12]. As part of the revised diagnostic criteria of PTSD, a newly formulated symptom cluster, entitled negative alterations in cognitions and mood (NACM), was included to reflect and highlight posttrauma changes in cognitions. In particular, NACM symptoms include changes in cognitions about the self (e.g. "I am inadequate."), others (e.g. "Everyone is malevolent."), and the world (e.g. "The world is a dangerous place."), as well as cognitions regarding the meaning of posttrauma symptoms (e.g. "I'm going crazy."). Research has provided strong support for the addition of cognitive symptoms to the diagnostic criteria for PTSD [13]. Confirmatory factor analyses conducted with two large surveys [1,14] indicated that the new four cluster model of PTSD (re-experiencing; avoidance; hyperarousal and reactivity; negative alterations in cognitions and mood) provided a better fit to the data than the previous three cluster model of PTSD that excluded NACM [14]. Moreover, cognitive distortions captured within the NACM cluster have been shown to predict the development, chronicity, and severity of PTSD, as well as functional impairment associated with PTSD symptoms [10,11,15,16].
Accordingly, a central component in many empirically supported treatments for PTSD is the targeting of trauma-related cognitions. In particular, evidence has shown belief change to be one mechanism by which cognitive-behavioral treatments for PTSD have led to reductions in symptom severity [17][18][19][20]. In one of the first outcome studies examining the efficacy of Cognitive Therapy for PTSD (CT-PTSD) it was observed that reduction in negative traumarelated cognitions resulted in a reduction in PTSD symptoms over the course of CT-PTSD in a treatment-seeking community sample, though it should be noted that the cell sizes for both the intervention and waitlist conditions were quite small (n = 14) [21]. While trauma-related cognitions are not specifically targeted for change in Prolonged Exposure (PE) [22], a first line treatment for PTSD that indirectly challenges trauma-related cognitions, a recent study with female survivors of assault enrolled in PE showed that changes in PTSD-related cognitions predicted subsequent changes in PTSD symptoms overall; PTSD symptom reduction did not predict change in cognitions [20]. Further, a recent longitudinal study found reductions in trauma-related cognitions were related to reductions in posttraumatic stress symptoms up to 10 years after treatment in a sample of 171 women randomized to either Prolonged Exposure or Cognitive Processing Therapy treatments for PTSD [23].
Currently, measures that tap into trauma-related cognitions include the Posttraumatic Cognitions Inventory (PTCI) [24]; World Assumptions Scales (WAS) [25], Trauma and Attachment Belief Scale (TABS) [26], Trauma-Related Guilt Inventory (TRGI) [27], Trauma Appraisal Questionnaire (TAQ) [28], and Posttraumatic Maladaptive Beliefs Scale (PMBS) [29]. These measures have contributed significantly to an understanding of trauma-related cognitions involved in the trajectory of, and recovery from, PTSD symptoms. Yet, the existing measures do not assess the full range of trauma cognitions associated with all three cognitive processes discussed. That is, existing measures generally provide either information on assimilated cognitions or overaccommodated cognitions. These tools are useful in identifying erroneous and extreme cognitions, yet they fail to assess accommodated cognitions. Accommodated cognitions reflect balanced thinking about the causes and consequences of the traumatic event and may be important in assessing changes associated with recovery in trauma-focused therapy.
The primary aim of the current study was to develop a measure of cognitions consistent with the cognitive processes involved in the development and maintenance of PTSD, and one that is theory-driven and aligned with evidence-based cognitive-behavioral trauma-focused therapies. Moreover, our goal was to take a first step in developing an assessment tool that would be useful in both research and clinical contexts. A comprehensive trauma cognitions measure that taps into all three cognitive processes would be beneficial to inform treatment planning and evaluate treatment effects over time. Because research has shown that exposure to interpersonal trauma (i.e., trauma committed at the hands of another human) is the strongest correlate of cognitive symptoms of PTSD [30], the present study used samples of interpersonal trauma survivors to develop the trauma cognitions measure across the final phases of the study.

Method
Scale construction was carried out across four phases. In the Item Generation Phase (Phase 1), the initial version of the Trauma-Related Cognitions Scale (TRCS) was created with items gathered from several validated and reliable trauma cognitions/beliefs measures in open access, as well as items generated by the first, second, and fourth author, who are all clinical research experts in trauma psychology and licensed clinical psychologists trained in traumafocused therapy. Permission to use items from pre-existing scales was secured by the original authors of those scales. In the Exploratory Factor Analysis Phase (Phase 2), the initial version of the TRCS was administered to a non-clinical sample and an exploratory factor analysis was fit to the data to explore the factor structure and reduce TRCS items. This study was approved by Northern Illinois University's Institutional Review Board and no adverse events were reported. In the Confirmatory Factor Analysis and Validity Assessment Phase (Phase 3), a confirmatory factor analysis (CFA) model was fit to the data from a second non-clinical trauma sample to confirm the empirical measurement model identified in Phase 2; criterion-related, convergent and discriminant, concurrent, and incremental validity were also examined. This study was approved by Northern Illinois University's Institutional Review Board and no adverse events were reported. In Phase 4, a CFA model was fit to data from a clinical sample presenting for trauma-focused therapy at an outpatient community mental health clinic, and validity was further assessed in this sample. This study was approved by the University of California, San Francisco medical center's Institutional Review Board and no adverse events were reported. present cognition (e.g., TAQ: I was a bad person was re-worded to I am a bad person) and to make items more concise.
From the pool of 94 items that were retained, the first, second, and fourth authors independently coded all items into one of three categories (1) overaccommodation, (2) assimilation, or (3) accommodation. The three authors then met to compare their independent codes and discuss any discrepancies until consensus was reached for which scale each item should be placed. Overaccommodation was represented by 46 items (27 items from the PTCI, 7 items from the TAQ, 12 items from the PMBS). Assimilation was represented by 38 items (15 items from the TRGI, 5 items from the PTCI, 9 items from the TAQ, 9 items from the WAS). Accommodation was represented by 10 items (7 items from the WAS, 3 items from the PMBS). After reading through item content for each subscale, it was agreed that additional items would be created to increase content validity. The first, second, and fourth authors independently created 8 provisional assimilation items and 87 provisional accommodation items. The higher number of accommodation items was created because few items from the existing scales were coded as representative of accommodative processes by the research team. After the creation of these items, the first, second, and fourth authors compared the independently created items to those from the 5 published instruments. Redundant items were discussed, and the authors came to consensus with the best wording for redundant and nonredundant items. Five items were added to the assimilation subscale and 35 items were added to the accommodation subscale. In total, the preliminary TRCS was represented by 134 items among the three subscales of overaccommodation (46 items), assimilation (43 items), and accommodation (45 items).
Finally, given the original five trauma cognitions measures had different response options, we standardized the TRCS anchors and chose six response options. Directions to complete the TRCS were modeled after the PTCI [24], and were as follows: We are interested in the kind of thoughts which you may have had after a traumatic experience. Below are a number of statements that may or may not be representative of your thinking.
Please read each statement carefully and tell us how much you AGREE or DISAGREE with each statement using the following rating scale: 1 (strongly disagree), 2 (disagree), 3 (somewhat disagree), 4 (somewhat agree), 5 (agree), 6 (strongly agree). People react to traumatic events in many different ways. There are no right or wrong answers to these statements. Instructions were kept consistent throughout all study phases.

Phase 2 method
Phase 2 participants and recruitment procedure. The Exploratory Factor Analysis Phase collected and analyzed TRCS data from a non-clinical sample. The sample included 815 participants: 509 (62.5%) community individuals recruited in 2013 through Amazon's Mechanical Turk (MTurk) who completed the study for small payments as well as 306 (37.5%) undergraduate students from a large Midwestern university who completed the study for course credit. Studies have found MTurk samples to be more socio-economically and ethnically diverse than participants recruited via social media postings and university settings, with indistinguishable test responses across samples that complete testing online and in-person. Thus, researchers have suggested that MTurk data is as reliable as those obtained via traditional methods [31,32].
Participants age 18 and above provided electronic consent to complete the TRCS, the Traumatic Life Events Questionnaire, and other measures described below online as part of a larger study. Following the completion of the study, participants were thanked for their participation and provided a debriefing form.
The average age of participants was 27.38 (SD = 10.30), ranging from 18 to 67. The majority of participants identified as male (54.3%) and identified their race as Caucasian/White (59.9%); 10.3% identified as African American, 21.7% identified as Asian, 1.1% identified as American Indian or Alaskan Native, 0.8% identified as Native Hawaiian/another Pacific Islander, 2.5% identified as Biracial, 0.8% reported Unknown, and 2.8% preferred not to answer; 10.4% identified their ethnicity as Hispanic or Latino. About a quarter of participants had a college degree (24.5%), 21% reported having some high school, 16.4% reported having a high school degree/GED, 17.5% reported having some college/vocational training, 2.0% reported having some graduate training, and 14.2% reported having a graduate degree. The majority were employed full-time (35.1%) or part-time (27.1%), and were single (45.8%) or dating (19.4%). The majority of participants had experienced a potentially traumatic event (see Table 1).
Phase 2 measures. Trauma cognitions. The original 134-item TRCS was used to assess trauma cognitions.
Trauma history. Potential traumatic experiences were identified using the Traumatic Life Events Questionnaire (TLEQ) [33], a 23-item broad-spectrum measure of trauma exposure. Items ask respondents to identify how many times they have experienced a particular event using the following rating scale: never (0), once (1), twice (2), 3 times (3), 4 times (4), 5 times (5), and more than 5 times (6). In one study utilizing a clinical sample, the TLEQ demonstrated a higher rate of identification of traumatic events compared to the Structured Clinical Interview for DSM-IV (SCID) [34]. Phase 2 data analysis plan. Exploratory factor analysis models with direct oblimin (oblique) rotation were fit to the respondent data on the 134-item initial TRCS. Data were assumed missing at random and missing data were deleted listwise. Factor solutions with varying numbers of factors (i.e., 2, 3, 4, and 5) were generated and the 4-factor solution was chosen as it offered a combination of theoretical appeal and parsimony. Items with maximum absolute factor pattern loadings below .30 and items that cross-loaded (loaded on two or more factors) were dropped if the larger factor loading was less than twice as large as the lesser loading [35]. Following EFA modeling, scale scores were calculated for each factor by averaging each respondent's corresponding item scores, such that scale scores had a possible range from 1-6. Inter-scale correlations and Cronbach alphas were estimated.

Phase 2 results
Phase 2 exploratory factor analysis. The chosen exploratory factor analysis model retained 69 items representing the three hypothesized factors of Overaccommodation (e.g., "I can't deal with even the slightest upset"; "My life has been destroyed by the trauma"; 25 items; α = .97), Assimilation (e.g., "I blame myself for what happened"; "What I did was inconsistent with my beliefs"; 13 items; α = .94), and Accommodation (e.g., "Life is sometimes a gamble"; "Life is about surviving challenging events"; 15 items, α = .89). In addition, an Optimism factor emerged that was not hypothesized, but made substantive sense (e.g., "The world is a good place"; "Most people are basically caring"; 16 items; α = .93; Table 2).
Between 10% and 20% of data values were missing. A missing data rate of 15% to 20% is common in psychological studies [36]. These four factors explained a total of 49.83% of total response variation. Scale score means, standard deviations, and inter-correlations are presented in Table 3. Item endorsement on each scale suggested participants 'disagreed' to 'somewhat disagreed' with items assessing overaccommodation; were ambivalent about items that assessed assimilation, with average item ratings ranging between 'somewhat disagree' to 'somewhat agree'; and generally agreed with items that assessed accommodation and optimism, with average item ratings between 'somewhat agree' and 'agree.' For each scale, internal consistency reliability was very good to excellent [37]. Overaccommodation was positively associated with assimilation and negatively associated with optimism; assimilation was positively associated with accommodation; and accommodation was positively associated with optimism (Table 3).

Phase 3 method
Phase 3 participants and recruitment procedure. The Phase 3 sample initially included 998 community individuals recruited through Amazon's MTurk in 2014. Research has found MTurk samples are more attentive to instructions than traditional subject pool samples [38]. Participants age 18 and above provided electronic consent to complete the measures online. Following the completion of the study, participants were thanked for their participation and provided a debriefing form. The majority of participants identified as male (51.2%) and identified as Caucasian/White (57.4%); 14.9% identified as African American, 18.9% identified as Asian, 1% identified as American Indian/Alaskan Native, 3.3% identified as biracial, and 2.2% reported 'unknown'; 11.5% identified their ethnicity as Hispanic or Latino. About a quarter of participants had a high school degree/GED (24.4%), 0.9% reported having some high school, 39.4% reported having some college/vocational training, 22.3% reported a college degree, 1.3% reported some graduate training, and 11.7% reported having a graduate degree. The majority were employed full-time (31.7%) or part-time (27.8%). Most participants had experienced a potentially traumatic event (see Table 1).
For psychometric analyses, the sample was restricted to those who reported experiencing at least one interpersonal trauma (i.e., an event in which another human being inflicts physical or psychological injury on another human being; n = 651). Interpersonal traumas were selected based upon TLEQ responses using the same procedure as Frazier and colleagues [39], except for abortion, which was not conceptualized in this study as an interpersonal trauma. Participants selected for interpersonal trauma, compared to those who were dropped from analyses, reported significantly more posttraumatic stress on the PTSD Checklist-Civilian Version (PCL-C; described below) [40], t(994) = 8.34, p < .001. Using the most stringent cutpoint score of 50, 25.8% would be considered PTSD positive in the interpersonal trauma subsample; whereas, 12.7% in the non-trauma sub-sample would be considered PTSD positive. Phase 3 measures. Trauma cognitions. The reduced 69-item TRCS from the previous phase was used.
Trauma history. The TLEQ was used to assess potential traumatic experiences. PTSD symptoms. The PTSD Checklist-Civilian Version (PCL-C) [40] was used to assess PTSD symptoms. The PCL-C is a 17-item self-report scale for posttraumatic stress based on DSM-IV criteria. Items are rated from 1 (Not at All) to 5 (Extremely). A total severity score is obtained by summing scores from each of the 17 items. The PCL-C has demonstrated high internal consistency ranging from .92 to .96 in community samples, and has shown convergent validity with other measures of posttraumatic stress and discriminant validity with measures of separate psychological constructs [41]. Internal consistency in the subsample selected for interpersonal trauma was α = .95.
Depressive symptoms. The Beck Depression Inventory-II (BDI-II) [42] is a 21-item selfreport measure for assessing the severity of depression. Items are scored from 0 to 3 with different anchors for each item (e.g., 1 = I do not feel sad to 3 = I am so sad or unhappy that I can't stand it), with higher scores indicating more severe depressive symptoms. The average testretest reliability is r = .93 and the average internal consistency is α = .92 [42]. In the Phase 3 sample, internal consistency was α = .94.
Resilience. The Brief Resilience Scale [43] is a 6-item measure that operationalizes resilience as bouncing back from stress ("It does not take long to recover from a stressful event"). Items  [43]. In the Phase 3 sample, internal consistency was α = .85. Personality. The Big Five Inventory [44] consists of 44 short-phrase items, rated on a fivepoint scale from 1 (Disagree Strongly) to 5 (Agree Strongly), that represent the core traits that define the Big Five personality domains of Extraversion ("Is outgoing, sociable"), Neuroticism ("Worries a lot"), Conscientiousness ("Does a thorough job"), Agreeableness ("Has a forgiving nature"), and Openness to Experience ("Is ingenious, a deep thinker"). Internal consistency in this sample ranged from .74 (Agreeableness) to .86 (Neuroticism).
Trait anxiety. The State-Trait Anxiety Inventory [45] was used to assess trait-anxiety, a relatively stable individual difference in anxiety proneness and the tendency to perceive stressful situations as dangerous or threatening. Trait anxiety was measured with 10 items ("I feel nervous and restless") on a 4-point scale ranging from 1 (Almost Never) to 4 (Almost Always). Internal consistency across different samples ranged from .89 to .91. Internal consistency in this sample was (α = .91).
Trauma beliefs. The TABS [26] is an 84-item self-report assessment of cognitive schemas (beliefs about oneself and about others). The TABS measures beliefs related to five need areas that are sensitive to the effects of traumatic experiences: safety, trust, esteem, intimacy, and control. Within each need area, separate sets of items tap into beliefs about oneself and beliefs about others. Items are rated from 1 (Disagree Strongly) to 6 (Agree Strongly), with higher scores indicating more negative beliefs about oneself and others. Internal consistency (.67 to .87) and test-retest reliabilities (.60 to .79) for the TABS subscales are adequate [26]. Internal consistency in this sample ranged from .85 (Intimacy) to .92 (Control).
Phase 3 data analysis plan. A confirmatory factor analysis (CFA) model was fit using LIS-REL 8.72 as a confirmatory test of the 69-item, four-factor TRCS measurement model empirically identified by the Phase 2 EFA model. Model fit was assessed by the Satorra-Bentler scaled chi-square (χ 2 SB ) [46], the root mean squared error of approximation (RMSEA) [47], and the comparative fit index (CFI) [48]. RMSEA values �.06 and CFI values � .95 are thought to suggest approximate model fit [49]. Across 69 items, 3.3% of data values were missing. To accommodate missing data the CFA model was fit to multiply imputed data and the asymptotic covariance matrix of item variances and covariances was estimated via the bootstrap. Following CFA modeling, scale scores were calculated for each factor. Inter-scale correlations and Cronbach alphas were estimated.

Phase 3 results
Phase 3 confirmatory factor analysis. The 4-factor, 69-item CFA model fit well: Satorra-Bentler χ 2 (2271) = 7394.72, p < .0001; RMSEA = 0.059; CFI = 0.967. See Table 2 for factor loadings. The four factors demonstrated high internal consistency reliability in the Phase 3 sample of non-clinical participants selected for interpersonal trauma exposure: Overaccommodation (α = .97), Assimilation (α = .93), Accommodation (α = .87), Optimism (α = .90). Scale score means, standard deviations, and inter-correlations are presented in Table 3. In this non-clinical sample selected for interpersonal trauma, it was found that item endorsement on each scale suggested participants 'disagreed' to 'somewhat disagreed' with items assessing overaccommodation; were ambivalent about items assessing assimilation, with average item ratings ranging between 'somewhat disagree' to somewhat agree'; and generally agreed with items assessing accommodation and optimism, with average items ratings between 'somewhat agree' and 'agree.' Again, similar to Phase 2 results, overaccommodation was highly positively associated with assimilation and negatively associated with optimism; assimilation was positively associated with accommodation; and accommodation was moderately positively associated with optimism.
Convergent and discriminant validity. See Table 4 for bivariate correlations between the four TRCS subscales and both trauma beliefs and personality factors.

Phase 4 method
Phase 4 participants and recruitment procedure. Participants included 73 acute victims of crime who were presenting for weekly trauma-focused therapy from 2015 to 2016 with one of 17 clinicians at a community mental health clinic in a large metropolitan area on the U.S. West Coast. Clinicians at the community mental health clinic approached potential clients during an intake evaluation about participating in the study, and those who expressed interest in participating met with a research assistant to learn more about the study. After providing written informed consent, participants were given a questionnaire packet that took approximately 45 minutes to complete. Demographic data, trauma history, and psychiatric diagnoses assessed at intake were extracted from clinical records. Following the completion of the packet of questionnaires, participants were thanked for their time and debriefed. They were provided a payment of $20 for their time.
The average age of participants was 41.28 (SD = 13.80), ranging from 21 to 87. The majority of participants identified as female (60.3%). Participants were racially diverse: 33.2% identified as Hispanic, 24.7% identified as Caucasian/White, 13.7% identified as African American/ Black, 9.6% identified as Asian/Pacific Islander, 1.4% identified as biracial, 15.1% specified 'Other' as their race, and 1.4% declined to provide a response. The majority of participants identified as heterosexual (58.9%); 13.7% identified as homosexual, 5.5% identified as bisexual, 21.9% either did not provide or declined to provide their sexual orientation.
Phase 4 measures and summary statistics. Trauma cognitions. In Phase 4, the reduced 69-item TRCS that resulted from the Phase 2 EFA and that was confirmed in the Phase 3 CFA was administered.
Trauma history. Trauma history was assessed in 63 participants with the Trauma History Screen (THS) [50] during an intake evaluation at the community mental health clinic before clients enrolled in trauma services at the clinic. The THS is a 14-item self-report measure that assesses exposure to a variety of traumatic events, 13 specific types of events (e.g., car accident, sexual assault, physical assault, natural disaster, etc.) and one "other" event. During the intake interview, for each event, participants were asked to indicate whether they experienced the event ("yes" or "no"). The average number of types of trauma experienced, as assessed by the THS, was 4.81 (SD = 2.93) and ranged up to 12 discrete traumas (Table 5).
To be eligible for trauma services at the clinic, the client had to be a victim of a crime within the last 3 years, or a survivor of torture in another country, or a family member of a homicide victim. The index trauma that precipitated mental health treatment at the clinic was identified for 64 participants during the intake evaluation. Most participants identified sexual assault (32.9%), followed by physical assault (26%); 9.6% were family members of homicide victims, 5.5% were refugees or victims of torture in their home country, 4.1% were victims of vehicular assault, 2.7% were victims of a stabbing, 2.7% were victims of attempted murder, 1.4% were victims of a shooting, and 2.7% of participants reported another type of crime as their index trauma.
Psychiatric diagnoses. Psychiatric diagnoses were assessed for 63 participants during an intake evaluation before clients enrolled in trauma services at the community mental health clinic by clinicians who conducted a clinical interview and utilized symptom checklists to identify psychiatric symptoms. After the intake, clinicians met weekly for an intake meeting to solidify psychiatric diagnoses made during the intake. Almost all of the sample (92.1%) met criteria for PTSD or another Trauma-and-Stressor-Related Disorder as defined by the DSM-5 [12]. The average number of diagnoses was 2.11 (SD = 1.44), ranging from zero (two participants did not meet criteria for any disorder) to eight (in the case of multiple substance use disorders), with a Depressive Disorder being the most common co-morbid diagnosis (39.7%). Other diagnoses included a Substance Use Disorder (19.1%), Anxiety Disorder (16.4%), Bipolar Disorder (4.8%), Psychotic Disorder (3.2%), Eating Disorder (3.2%), or Gender Dysphoria (1.6%).
PTSD symptoms. The PTSD Checklist-Civilian Version (PCL-C) [40] and the PTSD Checklist for DSM-5 (PCL-5) [51] were both used to assess PTSD symptoms. The PCL-5 is a 20-item self-report scale for posttraumatic stress based on DSM-5 criteria. Items are rated from 0 (Not at All) to 4 (Extremely). A total severity score is obtained by summing scores from each of the 20 items. Internal consistency for the PCL-5 has ranged from .91 (military service member sample seeking treatment) [52] to .94 (college sample) [53]. Part way through data collection, the clinic switched from using the PCL-C to the PCL-5 to assess posttraumatic stress as the new PCL version became available after the release of the DSM-5 [12]. Researchers at the clinic created a crosswalk between the two versions (i.e., a scoring rubric was created to produce scores that were equivalent across both versions of the measure). The PCL total is used as the

PLOS ONE
PTSD severity score in this Phase. The PCL was administered prior to initiating traumafocused therapy (baseline), as well as the beginning of sessions 8 and 16. Internal consistency for the PCL total in this sample ranged from .89 (baseline) to .93 (session 16). Depressive symptoms. The Patient Health Questionnaire-9 [54] is a 9-item self-report assessment that can be used as a continuous measure of depressive symptoms. Items ask participants to rate if they have been bothered by each symptom (e.g., "Little pleasure or interest in doing things") in the past 2 weeks on a scale from 0 (not at all) to 3 (nearly every day). A total severity score is obtained by summing the scores from each item. Internal consistency has ranged from .86 to .92 in several different samples [55]. The PHQ-9 was administered prior to initiating trauma-focused therapy (baseline), as well as the beginning of sessions 8 and 16. Internal consistency in this sample ranged from .79 (baseline) to .89 (therapy session 16).
Resilience. The Brief Resilience Scale [43] was used to assess resilience. In the Phase 4 sample, internal consistency was α = .84.
Trait anxiety. The State-Trait Anxiety Inventory [45] was used to assess trait-anxiety. Internal consistency in this sample was (α = .91).
Trauma beliefs. The TABS [26] assessed beliefs related to five need areas that are sensitive to the effects of traumatic experiences-safety, trust, esteem, intimacy, and control. Internal consistency in this sample ranged from .78 (Control) to .83 (Trust).
Phase 4 data analysis plan. Again, a four-factor confirmatory factor analysis (CFA) model was fit to the 69-item TRCS data. Across 69 items, 3.3% of data values were missing. Across 69 items, 0.54% of data values were missing. To accommodate missing data the CFA model was fit to multiply imputed data and the asymptotic covariance matrix of item variances and covariances was estimated via the bootstrap. Following CFA modeling, scale scores were calculated for each factor, and reliability and validity assessments conducted.
Item endorsement on each scale suggested participants 'disagreed' to 'somewhat disagreed' with items assessing overaccommodation; were ambivalent about items that assessed assimilation, with average item ratings ranging between 'somewhat disagree' to 'somewhat agree'; and generally agreed with items that assessed accommodation and optimism, as average item ratings were between 'somewhat agree' to 'agree' (see Table 6). Internal consistency reliability ranged from .62 (accommodation) to .92 (overaccommodation). Overaccommodation was positively and negatively associated with assimilation and optimism, respectively. However, accommodation was not significantly associated with any subscale.
Phase 4 criterion-related validity. Convergent and discriminant validity. See Table 7 for bivariate correlations between the four TRCS subscales and both trauma beliefs and personality factors.
Predictive validity. The PCL and the PHQ9 were assessed longitudinally. For each outcome, we fit five linear mixed effect longitudinal models. The first model included the 3-category assessment time indicator as the only X variable (baseline/pre-therapy, week 8, and week 16). This base model estimated the longitudinal effect of therapy on the corresponding outcome. Subsequently, four augmented models were fit; each included one TRCS scale as well as the interaction between the TRCS scale and the assessment time indicator. Non-significant interaction effects were dropped and the augmented model refit. Augmented models with no interaction term estimated the time-averaged effect of the modeled TRCS scale on the corresponding outcome. Models including an interaction term estimated time-specific effects of the modeled TRCS scale on the outcome. Each model included random effects for clinicians and patients and assumed data were missing at random (MAR), conditional on modeled variables.
Both the PCL and PHQ9 were completed by 73 (100%), 64 (87.7%), and 47 (64.4%) clients at the three assessments, respectively. The PCL means equaled 47.7, 27.7, and 22.8 at the three assessments. Base model tests of the baseline PCL mean versus the 8-and 16-week means were significant (both p < .001) and the 8-and 16-week means significantly differed (p = .036); this pattern also was observed in each augmented model. The interaction term was dropped from each augmented PCL model (p-values ranged from .217 to .826). Linear regression parameter estimates for the time-averaged effect of each TRCS scale on PCL scores follow: overaccommodation, b = 9.61, p < .001; assimilation, b = 0.03, p = .982; accommodation, b = 13.16, p < .001; and optimism, b = −8.72, p < .001. For example, a one-point increase on the TRCS overaccommodation scale corresponded to an expected 9.61-point increase on the PCL.
The PHQ9 means equaled 14.9, 7.4, and 7.6 at the three assessments. Base model tests of the baseline PHQ9 mean versus the 8-and 16-week means were significant (both p < .001), but the 8-and 16-week means did not significantly differ (p = .867); this pattern held in each augmented model. The interaction effect was dropped from three of the augmented PHQ9 models (p-values ranged from .667 to .984) and the linear regression parameter estimates for the corresponding time-averaged effects follow: overaccommodation, b = 3.90, p < .001; accommodation, b = 3.76, p = .004; and optimism, b = −8.72, p < .001. A significant interaction effect between assimilation and assessment time (p = .048) resulted in time-specific effects of assimilation on PHQ9: pre-therapy, b = 1.80, p = .008; 8 weeks, b = −0.08, p = .909; and 16 weeks b = 0.22, p = .785). The pre-therapy effect significantly differed from the 8-week effect (p = .021) and marginally differed from the 16-week effect (p = .071). Thus, assimilation was significantly related to pre-therapy PHQ9 scores, but that relationship no longer held by the 8-week therapy session.

Discussion
The aim of this research was to develop a measure that assesses cognitions consistent with cognitive processes of assimilation, accommodation, and overaccommodation. The Cognitive Theory of PTSD [56] asserts PTSD develops and persists in the aftermath of traumatic events that alter or reinforce negative cognitions about oneself, people, and the world. Cognitive Processing Therapy (CPT) [7] follows from this theory, arguing that PTSD symptomatology endures among individuals who assimilate (e.g., "I should not have gotten drunk that night") or over-accommodate (e.g., "All men are evil") the traumatic event into their core cognitive schemas about the world. A randomized clinical trial to dismantle the active components of CPT [57] demonstrated evidence for belief change as a mechanism of treatment success, highlighting the importance of identifying and eventually accommodating trauma-related cognitions into a schema system that recognizes the world as imperfect, but not wholly malevolent (e.g., "Sometimes people do bad things").
Despite evidence that post-trauma cognitions consistent with maladaptive informationprocessing play a role in PTSD onset, maintenance, and recovery, post trauma cognition measures used in research, treatment planning, and symptom monitoring typically reflect cognitions associated with just one cognitive process. Few measures incorporate items reflective of accommodated cognitions, which are items that would likely be most indicative of treatmentrelated success. The present study used two non-clinical samples and one clinical sample to develop a measure of cognitions consistent with all three cognitive processes, with an unanticipated fourth factor reflecting optimism emerging from analyses. Though preliminary, the measure has potential to impact the trauma field by providing empirical support for cognitive theories of PTSD, revealing the stability of post-trauma cognitions with and without intervention, and identifying idiosyncratic cognitions that may be targeted by clinicians providing trauma-focused treatment or cognitions that predict symptom trajectory.
Across the three samples, participants generally disagreed with items on overaccommodation, scored in the mid-range for assimilation items, and agreed with cognitions associated with accommodation and optimism. This may be anticipated in regard to Phase 3 results, as the majority of the participants did not endorse clinical levels of PTSD symptomatology despite exposure to interpersonal trauma. It is possible the overaccommodated and assimilated cognitions are more specific to PTSD development than trauma exposure in the absence of PTSD symptoms. However, the same pattern of results was found in the clinical sample. This may indicate that clinical samples hold idiosyncratic trauma cognitions that may not reflect average scores on the subscale. With this in mind, clinicians may use this measure as a tool to assess which idiosyncratic cognitions are most relevant to that individual and track those specific cognitions over therapy. Outcome research using the TRCS will shed light on whether the average scores found in this study replicate in subsequent clinical samples, as well as determine the clinical utility of the measure.
Contrary to expectations, a fourth factor emerged with items that reflect post-trauma optimism. Although unanticipated, the optimism domain makes theoretical sense, as some individuals may endorse cognitions about themselves and their recovery that are more purely positive in nature rather than the more neutral cognitions associated with accommodation or the more maladaptive cognitions associated with assimilation or over-accommodation. Though the optimistic cognitions factor emerged across all three subsamples, the extent to which the optimism factor has utility in research or clinical contexts remains an empirical question.
Several limitations to the current study must be noted. Phases 2 and 3 used Mechanical Turk to collect data, which may have increased selection effects [58]. As is common in measure development, the initial 4-factor CFA model fit to the clinical sample did not fit the data adequately, and thus the CFA model was empirically modified by dropping some items; yet, the number of factors and the groupings of remaining items within factors were unchanged. Therefore, the TRCS measurement model within clinical populations should be considered provisional until tested in subsequent clinical samples with a larger number of participants. In addition, the present study did not assess for several factors that have been implicated in posttrauma cognitions and reactions. For example, time elapsed since an index traumatic event was not assessed. Survivors' cognitions regarding an event may change naturally over time, or be altered with formal or informal intervention, the latter of which was also not assessed. Future research that considers how time elapsed since participants' index traumatic event, and previous exposure to intervention, impact cognitions assessed by the TRCS is needed. Further, in the non-clinical trauma sample in Phase 3 of the present study, the extent to which the potentially traumatic events assessed by the TLEQ resulted in traumatization (and therefore were truly traumatic) was not assessed. However, use of a cut-off score to examine the prevalence of probable PTSD in interpersonal trauma (IPT) survivors (versus those without such exposure) revealed nearly 25% of IPT survivors met cut-off for probable PTSD using a conservative cut point. As such, it is reasonable to assume that a sizable minority of the IPT survivors found their index event to be traumatizing. This limitation was largely rectified in the last phase, as 91% qualified for a diagnosis of PTSD and all participants were seeking traumafocused treatment. However, future studies using the TRCS would benefit from collecting additional samples of PTSD-positive participants, as well as assessing associations between subscales of the TRCS and symptoms of other disorders frequently observed in trauma survivors (i.e., depression, anxiety, substance use). Further, additional information regarding participants' trauma history is important to collect to ensure those indicating potential exposure have experienced an event that would reasonably impact trauma-related cognitions. In addition, items reflecting more general responses (i.e., "I'm not safe") should be investigated further to determine whether these are trauma-specific or specific to post-trauma mental health outcomes.
Validity analyses in Phase 4 revealed preliminary evidence of concurrent validity, incremental validity, and predictive validity. Yet, replication is needed to further establish the validity of the scale. Additionally, it is important to note the internal consistency of the accommodation scale in Phase 4 was α = .62, which was collected at intake for trauma-focused treatment. It is likely accommodated beliefs are less consistent, as survivors of trauma may have idiosyncratic cognitions based on the nature of their trauma. For instance, it may be that an individual who was assaulted by a stranger holds negative cognitions related to safety, but may have fewer negative cognitions related to self-blame.
The instructions given on the TRCS may have also impacted study results. The instructions request that respondents consider the items in relationship to "a traumatic experience" without explicit definition of what a traumatic event entails. Inclusion of a definition of trauma or reference to a specific type of trauma may better orient participants to a life event that is truly traumatic. At present, it cannot be determined whether participants across the phases of the present study were considering a traumatic event that would fall within the DSM-5 criteria for trauma when responding to TRCS items, or whether they were considering a life event that fails to achieve this criterion. Further, the instructions request that participants consider the degree to which items represent their thinking and agreement with the statements. It is possible for participants to have recurrent cognitions with which they do not agree. As such, future research that incorporates changes to the TRCS instructions or response scale may improve the interpretability and utility of findings.
The aim of the current study was to develop a comprehensive measure of post-trauma cognitions. The resulting scale may have utility in research and clinical contexts; however, the length of the scale may be a barrier to its use. Future research to identify items with the greatest utility that maintain scale validity is needed. Further, test-retest reliability was not assessed in the current study and should be an area for targeted investigation in future research. Information on the stability of posttraumatic cognitions with and without intervention would represent an important contribution to the traumatology field. As data from the first three phases of this study were collected prior to the release of the PCL-5 [51], additional research is needed to determine whether the scale maintains its association with PTSD symptoms using DSM-5 criteria in non-clinical samples. The PCL-5 was used in Phase 4 with the clinical sample.
A strength of this study was demonstration that the four-factor structure of the TRCS was replicated across different samples, including university students, community participants, and a treatment-seeking clinical sample. Despite study limitations, a provisional instrument of cognitions associated with varied post trauma cognitive processes was developed across several phases using both non-clinical and clinical samples. Incorporation of the TRCS in future survey research and clinical trials will determine with greater certainty its valid and reliable assessment of the four post trauma cognition factors identified in the current study, and of particular interest, whether changes in TRCS scores are associated with enhanced treatmentrelated outcomes. At the present time, trauma clinicians may consider employing TRCS for descriptive purposes as part of treatment planning and monitoring.